{\rtf1\ansi\ansicpg1252\uc1 \deff0\deflang1033\deflangfe1033{\fonttbl{\f0\froman\fcharset0\fprq2{\*\panose 02020603050405020304}Times New Roman;}{\f1\fswiss\fcharset0\fprq2{\*\panose 020b0604020202020204}Arial;}
{\f2\fmodern\fcharset0\fprq1{\*\panose 02070309020205020404}Courier New;}{\f3\froman\fcharset2\fprq2{\*\panose 05050102010706020507}Symbol;}{\f4\froman\fcharset0\fprq2{\*\panose 02020603050405020304}Times{\*\falt Times New Roman};}
{\f5\fswiss\fcharset0\fprq2{\*\panose 020b0604020202020204}Helvetica;}{\f6\fmodern\fcharset0\fprq1{\*\panose 00000000000000000000}Courier{\*\falt Courier New};}{\f7\fswiss\fcharset0\fprq2{\*\panose 00000000000000000000}Geneva;}
{\f8\froman\fcharset0\fprq2{\*\panose 00000000000000000000}Tms Rmn{\*\falt Times New Roman};}{\f9\fswiss\fcharset0\fprq2{\*\panose 00000000000000000000}Helv;}{\f10\froman\fcharset0\fprq2{\*\panose 00000000000000000000}MS Serif;}
{\f11\fswiss\fcharset0\fprq2{\*\panose 00000000000000000000}MS Sans Serif;}{\f12\froman\fcharset0\fprq2{\*\panose 00000000000000000000}New York{\*\falt Times New Roman};}{\f13\fswiss\fcharset0\fprq2{\*\panose 00000000000000000000}System;}
{\f14\fnil\fcharset2\fprq2{\*\panose 05000000000000000000}Wingdings;}{\f15\fswiss\fcharset0\fprq2{\*\panose 020b0604030504040204}Tahoma;}{\f16\fswiss\fcharset0\fprq2{\*\panose 020b0604030504040204}Verdana;}
{\f17\fswiss\fcharset0\fprq2{\*\panose 00000000000000000000}AvantGarde;}{\f18\froman\fcharset0\fprq2{\*\panose 00000000000000000000}ITC Bookman;}{\f19\fswiss\fcharset0\fprq2{\*\panose 00000000000000000000}Helvetica-Narrow;}
{\f20\froman\fcharset0\fprq2{\*\panose 00000000000000000000}NewCenturySchlbk;}{\f21\froman\fcharset0\fprq2{\*\panose 00000000000000000000}Palatino;}{\f22\froman\fcharset0\fprq2{\*\panose 00000000000000000000}ZapfChancery;}
{\f23\fdecor\fcharset2\fprq2{\*\panose 00000000000000000000}ZapfDingbats;}{\f24\fnil\fcharset2\fprq2{\*\panose 00000000000000000000}Marlett;}{\f25\fmodern\fcharset0\fprq1{\*\panose 020b0609040504020204}Lucida Console;}
{\f26\fswiss\fcharset0\fprq2{\*\panose 020b0602030504020204}Lucida Sans Unicode;}{\f27\fswiss\fcharset0\fprq2{\*\panose 020b0a04020102020204}Arial Black;}{\f28\fswiss\fcharset0\fprq2{\*\panose 020b0506020202030204}Arial Narrow;}
{\f29\froman\fcharset0\fprq2{\*\panose 02050604050505020204}Bookman Old Style;}{\f30\froman\fcharset2\fprq2{\*\panose 05050102010706020507}Bookshelf Symbol 3;}{\f31\froman\fcharset0\fprq2{\*\panose 02020404030301010803}Garamond;}
{\f32\fswiss\fcharset0\fprq2{\*\panose 020b0706040902060204}Haettenschweiler;}{\f33\fswiss\fcharset0\fprq2{\*\panose 020b0806030902050204}Impact;}{\f34\froman\fcharset0\fprq2{\*\panose 00050102010706020507}Map Symbols;}
{\f35\fnil\fcharset2\fprq2{\*\panose 01010601010101010101}Monotype Sorts;}{\f36\froman\fcharset2\fprq2{\*\panose 05050102010205020202}MT Extra;}{\f37\fscript\fcharset0\fprq2{\*\panose 030f0702030302020204}Comic Sans MS;}
{\f38\fnil\fcharset2\fprq2{\*\panose 05000000000000000000}MS Outlook;}{\f39\fnil\fcharset2\fprq2{\*\panose 00000400000000000000}Fonetica;}{\f40\froman\fcharset2\fprq2{\*\panose 05030102010509060703}Webdings;}
{\f41\froman\fcharset0\fprq2{\*\panose 02040502050405020303}Georgia;}{\f42\fswiss\fcharset0\fprq2{\*\panose 020b0603020202020204}Trebuchet MS;}{\f43\fmodern\fcharset0\fprq1{\*\panose 020b0509000000000004}Andale Mono;}
{\f44\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Monaco;}{\f45\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Zapf Dingbats;}{\f46\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Charcoal;}
{\f47\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Myriad Roman;}{\f48\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Myriad Bold;}{\f49\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Myriad Italic;}
{\f50\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Myriad BoldItalic;}{\f51\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Myriad Tilt;}{\f52\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Letter Gothic MT Bold Oblique;}
{\f53\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Letter Gothic MT Bold;}{\f54\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Letter Gothic MT Oblique;}{\f55\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Letter Gothic MT;}
{\f56\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Monotype.com;}{\f57\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Times New Roman MT Extra Bold;}{\f58\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Klang MT;}
{\f59\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Script MT Bold;}{\f60\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Old English Text MT;}{\f61\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}New Berolina MT;}
{\f62\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Bodoni MT Ultra Bold;}{\f63\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Arial MT Condensed Light;}{\f64\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Gothic Bold Italic;}
{\f65\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Gothic Bold;}{\f66\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Gothic Italic;}{\f67\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Gothic;}
{\f68\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Bookman Old Style Bold Italic;}{\f69\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Bookman Old Style Bold;}
{\f70\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Bookman Old Style Italic;}{\f71\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Book Antiqua Bold Italic;}{\f72\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Book Antiqua Bold;}
{\f73\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Book Antiqua Italic;}{\f74\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Book Antiqua;}{\f75\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Schoolbook Bold Italic;}
{\f76\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Schoolbook Bold;}{\f77\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Schoolbook Italic;}{\f78\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Century Schoolbook;}
{\f79\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Arial Narrow Bold Italic;}{\f80\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Arial Narrow Bold;}{\f81\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Arial Narrow Italic;}
{\f82\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Gill Sans Condensed Bold;}{\f83\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Minion Web;}{\f84\fnil\fcharset77\fprq0{\*\panose 00000000000000000000}Chicago;}
{\f85\froman\fcharset0\fprq0{\*\panose 00000000000000000000}Bookman;}{\f86\froman\fcharset0\fprq2{\*\panose 00000000000000000000}Garamond MT{\*\falt Garamond};}{\f87\fnil\fcharset134\fprq2{\*\panose 00000000000000000000}SimSun{\*\falt ??};}
{\f88\froman\fcharset0\fprq0{\*\panose 00000000000000000000}??;}{\f89\fdecor\fcharset0\fprq2{\*\panose 04020805060202030203}Colonna MT{\*\falt Courier New};}{\f90\froman\fcharset0\fprq2{\*\panose 0204060206030a020304}Footlight MT Light;}
{\f91\froman\fcharset238\fprq2{\*\panose 00000000000000000000}TIMES CE{\*\falt Times New Roman};}{\f92\froman\fcharset204\fprq2{\*\panose 00000000000000000000}TIMES Cyr{\*\falt Times New Roman};}
{\f93\froman\fcharset161\fprq2{\*\panose 00000000000000000000}TIMES Greek{\*\falt Times New Roman};}{\f94\froman\fcharset162\fprq2{\*\panose 00000000000000000000}TIMES Tur{\*\falt Times New Roman};}
{\f95\froman\fcharset186\fprq2{\*\panose 00000000000000000000}TIMES Baltic{\*\falt Times New Roman};}{\f96\froman\fcharset0\fprq2{\*\panose 00000000000000000000}CG Times (W1){\*\falt Times New Roman};}
{\f97\froman\fcharset0\fprq2{\*\panose 00000000000000000000}CG Times (WN);}{\f98\froman\fcharset238\fprq2 Times New Roman CE;}{\f99\froman\fcharset204\fprq2 Times New Roman Cyr;}{\f101\froman\fcharset161\fprq2 Times New Roman Greek;}
{\f102\froman\fcharset162\fprq2 Times New Roman Tur;}{\f103\froman\fcharset186\fprq2 Times New Roman Baltic;}{\f104\fswiss\fcharset238\fprq2 Arial CE;}{\f105\fswiss\fcharset204\fprq2 Arial Cyr;}{\f107\fswiss\fcharset161\fprq2 Arial Greek;}
{\f108\fswiss\fcharset162\fprq2 Arial Tur;}{\f109\fswiss\fcharset186\fprq2 Arial Baltic;}{\f110\fmodern\fcharset238\fprq1 Courier New CE;}{\f111\fmodern\fcharset204\fprq1 Courier New Cyr;}{\f113\fmodern\fcharset161\fprq1 Courier New Greek;}
{\f114\fmodern\fcharset162\fprq1 Courier New Tur;}{\f115\fmodern\fcharset186\fprq1 Courier New Baltic;}{\f188\fswiss\fcharset238\fprq2 Tahoma CE;}{\f189\fswiss\fcharset204\fprq2 Tahoma Cyr;}{\f191\fswiss\fcharset161\fprq2 Tahoma Greek;}
{\f192\fswiss\fcharset162\fprq2 Tahoma Tur;}{\f193\fswiss\fcharset186\fprq2 Tahoma Baltic;}{\f194\fswiss\fcharset238\fprq2 Verdana CE;}{\f195\fswiss\fcharset204\fprq2 Verdana Cyr;}{\f197\fswiss\fcharset161\fprq2 Verdana Greek;}
{\f198\fswiss\fcharset162\fprq2 Verdana Tur;}{\f199\fswiss\fcharset186\fprq2 Verdana Baltic;}{\f248\fmodern\fcharset238\fprq1 Lucida Console CE;}{\f249\fmodern\fcharset204\fprq1 Lucida Console Cyr;}{\f251\fmodern\fcharset161\fprq1 Lucida Console Greek;}
{\f252\fmodern\fcharset162\fprq1 Lucida Console Tur;}{\f254\fswiss\fcharset238\fprq2 Lucida Sans Unicode CE;}{\f255\fswiss\fcharset204\fprq2 Lucida Sans Unicode Cyr;}{\f257\fswiss\fcharset161\fprq2 Lucida Sans Unicode Greek;}
{\f258\fswiss\fcharset162\fprq2 Lucida Sans Unicode Tur;}{\f260\fswiss\fcharset238\fprq2 Arial Black CE;}{\f261\fswiss\fcharset204\fprq2 Arial Black Cyr;}{\f263\fswiss\fcharset161\fprq2 Arial Black Greek;}{\f264\fswiss\fcharset162\fprq2 Arial Black Tur;}
{\f265\fswiss\fcharset186\fprq2 Arial Black Baltic;}{\f266\fswiss\fcharset238\fprq2 Arial Narrow CE;}{\f267\fswiss\fcharset204\fprq2 Arial Narrow Cyr;}{\f269\fswiss\fcharset161\fprq2 Arial Narrow Greek;}{\f270\fswiss\fcharset162\fprq2 Arial Narrow Tur;}
{\f271\fswiss\fcharset186\fprq2 Arial Narrow Baltic;}{\f272\froman\fcharset238\fprq2 Bookman Old Style CE;}{\f273\froman\fcharset204\fprq2 Bookman Old Style Cyr;}{\f275\froman\fcharset161\fprq2 Bookman Old Style Greek;}
{\f276\froman\fcharset162\fprq2 Bookman Old Style Tur;}{\f277\froman\fcharset186\fprq2 Bookman Old Style Baltic;}{\f284\froman\fcharset238\fprq2 Garamond CE;}{\f285\froman\fcharset204\fprq2 Garamond Cyr;}{\f287\froman\fcharset161\fprq2 Garamond Greek;}
{\f288\froman\fcharset162\fprq2 Garamond Tur;}{\f289\froman\fcharset186\fprq2 Garamond Baltic;}{\f296\fswiss\fcharset238\fprq2 Impact CE;}{\f297\fswiss\fcharset204\fprq2 Impact Cyr;}{\f299\fswiss\fcharset161\fprq2 Impact Greek;}
{\f300\fswiss\fcharset162\fprq2 Impact Tur;}{\f301\fswiss\fcharset186\fprq2 Impact Baltic;}{\f320\fscript\fcharset238\fprq2 Comic Sans MS CE;}{\f321\fscript\fcharset204\fprq2 Comic Sans MS Cyr;}{\f323\fscript\fcharset161\fprq2 Comic Sans MS Greek;}
{\f324\fscript\fcharset162\fprq2 Comic Sans MS Tur;}{\f325\fscript\fcharset186\fprq2 Comic Sans MS Baltic;}{\f344\froman\fcharset238\fprq2 Georgia CE;}{\f345\froman\fcharset204\fprq2 Georgia Cyr;}{\f347\froman\fcharset161\fprq2 Georgia Greek;}
{\f348\froman\fcharset162\fprq2 Georgia Tur;}{\f349\froman\fcharset186\fprq2 Georgia Baltic;}{\f350\fswiss\fcharset238\fprq2 Trebuchet MS CE;}{\f354\fswiss\fcharset162\fprq2 Trebuchet MS Tur;}{\f356\fmodern\fcharset238\fprq1 Andale Mono CE;}
{\f357\fmodern\fcharset204\fprq1 Andale Mono Cyr;}{\f359\fmodern\fcharset161\fprq1 Andale Mono Greek;}{\f360\fmodern\fcharset162\fprq1 Andale Mono Tur;}{\f361\fmodern\fcharset186\fprq1 Andale Mono Baltic;}{\f611\froman\fcharset161\fprq0 Bookman Greek;}
{\f613\froman\fcharset186\fprq0 Bookman Baltic;}}{\colortbl;\red0\green0\blue0;\red0\green0\blue255;\red0\green255\blue255;\red0\green255\blue0;\red255\green0\blue255;\red255\green0\blue0;\red255\green255\blue0;\red255\green255\blue255;
\red0\green0\blue128;\red0\green128\blue128;\red0\green128\blue0;\red128\green0\blue128;\red128\green0\blue0;\red128\green128\blue0;\red128\green128\blue128;\red192\green192\blue192;}{\stylesheet{\widctlpar\adjustright \lang2057\cgrid \snext0 Normal;}{
\s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec }\ls21\adjustright \b\fs28\cgrid \sbasedon0 \snext0 heading 1;}{\s2\sb240\sa60\keepn\widctlpar\adjustright \b\f4\fs28\lang2057\cgrid \sbasedon0 \snext0 
heading 2;}{\s3\qc\keepn\widctlpar\adjustright \b\lang2057\cgrid \sbasedon0 \snext0 heading 3;}{\s4\qc\keepn\widctlpar\adjustright \fs28\lang2057\cgrid \sbasedon0 \snext0 heading 4;}{\*\cs10 \additive Default Paragraph Font;}{\s15\widctlpar\adjustright 
\f6\fs20\lang2057\cgrid \sbasedon0 \snext15 Body Text;}{\s16\qj\fi709\widctlpar\adjustright \cgrid \sbasedon0 \snext16 \sautoupd par normal;}{\*\cs17 \additive \super \sbasedon10 footnote reference;}{\*\cs18 \additive \i\nosupersub \sbasedon10 capLabel;}{
\s19\qc\sb120\sa120\widctlpar\adjustright \fs20\cgrid \sbasedon0 \snext19 tabela;}{\s20\qj\widctlpar\adjustright \fs20\cgrid \sbasedon0 \snext20 footnote text;}{\s21\qc\sa120\sl-220\slmult0\widctlpar\adjustright \i\fs20\cgrid \sbasedon0 \snext0 caption;}
{\*\cs22 \additive \ul\cf2 \sbasedon10 Hyperlink;}{\s23\widctlpar\tqc\tx4536\tqr\tx9072\adjustright \lang2057\cgrid \sbasedon0 \snext23 footer;}{\*\cs24 \additive \sbasedon10 page number;}{\s25\qj\widctlpar\adjustright \fs20\lang2057\cgrid 
\sbasedon0 \snext25 endnote text;}{\*\cs26 \additive \super \sbasedon10 endnote reference;}{\s27\fi-567\li567\widctlpar\adjustright \cgrid \sbasedon0 \snext27 Body Text 2;}{\s28\qj\fi-567\li567\ri227\sa120\widctlpar\tx993\adjustright \f4\lang1044\cgrid 
\sbasedon0 \snext28 biblio;}{\s29\qj\sb240\sa120\sl360\slmult1\widctlpar\adjustright \b\fs28\lang2070\cgrid \sbasedon0 \snext29 titulo;}{\s30\qj\widctlpar\adjustright \fs20\lang2057\cgrid \sbasedon0 \snext30 Body Text 2;}{\*\cs31 \additive \ul\cf12 
\sbasedon10 FollowedHyperlink;}{\s32\widctlpar\tqc\tx4320\tqr\tx8640\adjustright \lang2057\cgrid \sbasedon0 \snext32 header;}{\s33\qj\widctlpar\adjustright \fs20\cgrid \sbasedon16 \snext33 exemplo;}{\s34\qj\widctlpar\adjustright \cgrid 
\sbasedon16 \snext34 \sautoupd prim par;}{\s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec }\ls24\adjustright \fs22\cgrid \sbasedon16 \snext35 \sautoupd exemplo numerado;}{\s36\qj\fi-360\li360\widctlpar
\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls27\pnrnot0\pndec }\ls27\adjustright \cgrid \sbasedon34 \snext36 lista;}{\s37\qj\fi709\sa120\widctlpar\adjustright \cgrid \sbasedon16 \snext37 par antes de lista;}}{\*\listtable{\list\listtemplateid878058312
\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li1492\jclisttab\tx1492 }{\listname ;}\listid-132}{\list\listtemplateid-825868584\listsimple{\listlevel
\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li1209\jclisttab\tx1209 }{\listname ;}\listid-131}{\list\listtemplateid-419940898\listsimple{\listlevel\levelnfc0\leveljc0
\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li926\jclisttab\tx926 }{\listname ;}\listid-130}{\list\listtemplateid-643652840\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1
\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li643\jclisttab\tx643 }{\listname ;}\listid-129}{\list\listtemplateid-191599476\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0
{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li1492\jclisttab\tx1492 }{\listname ;}\listid-128}{\list\listtemplateid-901509348\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext
\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li1209\jclisttab\tx1209 }{\listname ;}\listid-127}{\list\listtemplateid825792960\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext
\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li926\jclisttab\tx926 }{\listname ;}\listid-126}{\list\listtemplateid895882470\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext
\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li643\jclisttab\tx643 }{\listname ;}\listid-125}{\list\listtemplateid525228022\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext
\'02\'00.;}{\levelnumbers\'01;}\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid-120}{\list\listtemplateid32693612\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}
\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid-119}{\list\listtemplateid67698689\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 
\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid257561024}{\list\listtemplateid67698689\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360
\jclisttab\tx360 }{\listname ;}\listid343551711}{\list\listtemplateid67698689\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }
{\listname ;}\listid384718094}{\list\listtemplateid67698703\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid386614399}
{\list\listtemplateid854387670\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'03(\'00);}{\levelnumbers\'02;}\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid812715038}
{\list\listtemplateid-401434670\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \s36\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid972708565}
{\list\listtemplateid67698703\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1258714595}
{\list\listtemplateid1405886750\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fbias0 \s1\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1324888874}
{\list\listtemplateid1663893488\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'03(\'00);}{\levelnumbers\'02;}\fbias0 \s35\fi-720\li720\jclisttab\tx720 }{\listname ;}\listid1395815460}
{\list\listtemplateid67698703\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1422532447}{\list\listtemplateid67698703
\listsimple{\listlevel\levelnfc0\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'02\'00.;}{\levelnumbers\'01;}\fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1448041172}{\list\listtemplateid67698689\listsimple{\listlevel
\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1453091075}{\list\listtemplateid68419585\listsimple{\listlevel\levelnfc23
\leveljc0\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1493986829}{\list\listtemplateid67698689\listsimple{\listlevel\levelnfc23\leveljc0
\levelfollow0\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1577979394}{\list\listtemplateid68419585\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0
\levelstartat1\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1661039671}{\list\listtemplateid67698689\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1
\levelspace0\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1789816737}{\list\listtemplateid67698689\listsimple{\listlevel\levelnfc23\leveljc0\levelfollow0\levelstartat1\levelspace0
\levelindent0{\leveltext\'01\u-3913 ?;}{\levelnumbers;}\f3\fbias0 \fi-360\li360\jclisttab\tx360 }{\listname ;}\listid1980185603}}{\*\listoverridetable{\listoverride\listid1789816737\listoverridecount0\ls1}{\listoverride\listid1661039671
\listoverridecount0\ls2}{\listoverride\listid-120\listoverridecount0\ls3}{\listoverride\listid-129\listoverridecount0\ls4}{\listoverride\listid-130\listoverridecount0\ls5}{\listoverride\listid-131\listoverridecount0\ls6}{\listoverride\listid-132
\listoverridecount0\ls7}{\listoverride\listid-119\listoverridecount0\ls8}{\listoverride\listid-125\listoverridecount0\ls9}{\listoverride\listid-126\listoverridecount0\ls10}{\listoverride\listid-127\listoverridecount0\ls11}{\listoverride\listid-128
\listoverridecount0\ls12}{\listoverride\listid257561024\listoverridecount0\ls13}{\listoverride\listid812715038\listoverridecount0\ls14}{\listoverride\listid1422532447\listoverridecount0\ls15}{\listoverride\listid386614399\listoverridecount0\ls16}
{\listoverride\listid1980185603\listoverridecount0\ls17}{\listoverride\listid1453091075\listoverridecount0\ls18}{\listoverride\listid1448041172\listoverridecount0\ls19}{\listoverride\listid1258714595\listoverridecount0\ls20}{\listoverride\listid1324888874
\listoverridecount0\ls21}{\listoverride\listid384718094\listoverridecount0\ls22}{\listoverride\listid1577979394\listoverridecount0\ls23}{\listoverride\listid1395815460\listoverridecount0\ls24}{\listoverride\listid343551711\listoverridecount0\ls25}
{\listoverride\listid1493986829\listoverridecount0\ls26}{\listoverride\listid972708565\listoverridecount0\ls27}}{\info{\title An evaluation of the Translation Corpus Aligner with special reference to the language pair English-Portuguese}
{\author Signe Oksefjell}{\operator SINTEF}{\creatim\yr2000\mo7\dy31\hr16\min7}{\revtim\yr2000\mo7\dy31\hr16\min7}{\printim\yr2000\mo4\dy13\hr10\min29}{\version2}{\edmins0}{\nofpages4}{\nofwords5851}{\nofchars33353}{\*\company HiA}{\nofcharsws40959}
{\vern89}}\paperw11906\paperh16838\margl1701\margr1701\margt1701\margb1701 \deftab708\widowctrl\endnotes\aendnotes\hyphhotz425\makebackup\aftnnar\formshade\viewkind4\viewscale100\pgbrdrhead\pgbrdrfoot \fet1{\*\aftnsep \pard\plain 
\sb120\sa120\widctlpar\adjustright \lang2057\cgrid {\b\lang2070 Notes
\par }}{\*\aftnsepc \pard\plain \widctlpar\adjustright \lang2057\cgrid {\lang2070 
\par }}{\*\aftncn \pard\plain \widctlpar\adjustright \lang2057\cgrid {
\par }}\sectd \linex0\headery709\footery709\colsx709\endnhere\sectdefaultcl {\footer \pard\plain \s23\widctlpar\tqc\tx4536\tqr\tx9072\pvpara\phmrg\posxr\posy0\adjustright \lang2057\cgrid {\cs24 
\par }\pard \s23\ri360\widctlpar\tqc\tx4536\tqr\tx9072\adjustright {
\par }}{\*\pnseclvl1\pnucrm\pnstart1\pnindent720\pnhang{\pntxta .}}{\*\pnseclvl2\pnucltr\pnstart1\pnindent720\pnhang{\pntxta .}}{\*\pnseclvl3\pndec\pnstart1\pnindent720\pnhang{\pntxta .}}{\*\pnseclvl4\pnlcltr\pnstart1\pnindent720\pnhang{\pntxta )}}
{\*\pnseclvl5\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl6\pnlcltr\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl7\pnlcrm\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl8
\pnlcltr\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}{\*\pnseclvl9\pnlcrm\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\pard\plain \s29\qc\sb240\sa120\sl360\slmult1\widctlpar\adjustright \b\fs28\lang2070\cgrid {\fs32 
An evaluation of the Translation Corpus Aligner, with special reference to the language pair English-Portuguese
\par }\pard\plain \s4\qc\keepn\widctlpar\outlinelevel3\adjustright \fs28\lang2057\cgrid {Diana Santos & Signe Oksefjell
\par }\pard\plain \qc\widctlpar\adjustright \lang2057\cgrid {\fs28 SINTEF Telecom and Informatics & IBA, University of Oslo
\par }{\field\flddirty{\*\fldinst { HYPERLINK mailto:Diana.Santos@informatics.sintef.no }{{\*\datafield 
00d0c9ea79f9bace118c8200aa004ba90b0200000017000000230000004400690061006e0061002e00530061006e0074006f007300400069006e0066006f0072006d00610074006900630073002e00730069006e007400650066002e006e006f000000e0c9ea79f9bace118c8200aa004ba90b540000006d00610069006c00
74006f003a004400690061006e0061002e00530061006e0074006f007300400069006e0066006f0072006d00610074006900630073002e00730069006e007400650066002e006e006f000000}}}{\fldrslt {\cs22\ul\cf2 Diana.Santos@inf{\*\bkmkstart _Hlt470601379}o{\*\bkmkend _Hlt470601379}
rmatics.sintef.no}}}{ & }{\field\flddirty{\*\fldinst {\lang2070\cgrid0  HYPERLINK mailto:Signe.Oksefjell@hia.no }{\lang2070\cgrid0 {\*\datafield 
00d0c9ea79f9bace118c8200aa004ba90b0200000017000000170000005300690067006e0065002e004f006b007300650066006a0065006c006c0040006800690061002e006e006f000000e0c9ea79f9bace118c8200aa004ba90b3c0000006d00610069006c0074006f003a005300690067006e0065002e004f006b007300
650066006a0065006c006c0040006800690061002e006e006f000000}}}{\fldrslt {\cs22\ul\cf2 Signe.Oksefjell@hia.no}}}{\lang2070\cgrid0 
\par }{\fs28 
\par }\pard\plain \s3\qc\keepn\widctlpar\outlinelevel2\adjustright \b\lang2057\cgrid {Abstract
\par }\pard\plain \s30\qj\li113\ri113\widctlpar\adjustright \fs20\lang2057\cgrid {In this paper we describe the evaluation of a language-dependent aligner. We begin by introducing the alignment program, explaining why it w
ould be interesting to evaluate it with particular emphasis on the language pair English-Portuguese. A short presentation of the corpus used to test the aligner is also given. We then describe three experiments that were performed in the evaluation proces
s, presenting the results and discussing the methodology. The paper ends with a discussion of more general conclusions relative to an evaluation of this kind. 
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 1.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid {Introduction
\par }\pard\plain \s34\qj\widctlpar\adjustright \cgrid {Two criteria that are often employed in the evaluation of NLP programs are performan
ce and usability. Another criterion, less frequently mentioned, is the adequacy of handling particular languages. The present study describes a set of experiments devised to perform such an evaluation. 
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {Although researchers concerned with parallel corpus b
uilding and exploration will generally be happy to use a system available for their languages without evaluating it thoroughly, especially when the system is freely distributed \endash 
 as is the case of the present system, the kind of work reported originates fr
om two relevant concerns. The first one is about methodological aspects related to the development of NLP systems. The second concern is evaluation and comparison of products. In fact, there is a blatant lack of serious evaluation work of products and sys
tems concerning the Portuguese language, which is a situation we have been trying to change in the project Computational Processing of Portuguese at SINTEF.}{\cs26\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar\adjustright 
\fs20\lang2057\cgrid {\cs26\super \chftn }{ See http://www.portugues.mct.pt/.}}}{ 
\par The Translation Corpus Aligner (TCA) was developed in connection with the English-Norwegian Parall
el Corpus (ENPC) project with the aim of automatically aligning English and Norwegian texts (see e.g. Hofland 1996, Hofland & Johansson 1998). Although the program was originally written for the language pair English-Norwegian, it has been further develop
ed to handle other language pairs, including English-Portuguese. It includes a language-dependent component in the form of an anchor word list.
\par In the present paper we set out to evaluate the TCA for the language pair English-Portuguese. In particular, we want to
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
investigate the effect of the anchor word list; 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
compare the results of the program with and without the anchor word list; 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
find out how much a proofreader has to check manually after alignment
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {In order to perform the evaluation, we used the English-Portuguese part of the ENPC, which currently includes 16 English texts, about 12,000 words each, that have translations into Portuguese.}{
\cs26\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{
 One of the English texts has been counted twice since two translations of it have been included in the corpus; one into European Portuguese and one into Brazilian Portuguese. For a complete 
list of authors and texts in the corpus see the ENPC homepage at }{\field\flddirty{\*\fldinst { HYPERLINK http://www.hf.uio.no/iba/prosjekt/ }{{\*\datafield 
00d0c9ea79f9bace118c8200aa004ba90b02000000170000002300000068007400740070003a002f002f007700770077002e00680066002e00750069006f002e006e006f002f006900620061002f00700072006f0073006a0065006b0074002f000000e0c9ea79f9bace118c8200aa004ba90b460000006800740074007000
3a002f002f007700770077002e00680066002e00750069006f002e006e006f002f006900620061002f00700072006f0073006a0065006b0074002f000000}}}{\fldrslt {\cs22\ul\cf2 http://www.hf.uio.no/iba/prosjekt/}}}{
. For a more detailed description of the ENPC, see Johansson et al. (1999).}}}{ 
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 2.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid {A short description of the TCA
\par }\pard\plain \s34\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {The alignment program automatically matches original sentences with their translation
s. In the process of linking corresponding sentences, the program makes use of an anchor word list that contains word pairs of the languages in question. The one used here was not originally made for Portuguese and English, but was adapted from the Englis
h
-Norwegian anchor word list. In the alignment process, a value is given to the combination of sentences based on matches in the word list. The program goes through the texts and reads chunks of sentences in each language, resulting in matrices in which th
e
 program seeks the highest values for a match between sentences. In addition to the anchor word list, these values depend on the number of characters within the s-units/sentences in each language, on special characters (such as ?!;, etc.), on proper names
,
 and on cognates. One of the strengths of the program is that it does not require any preprocessing in the form of "hard regions", e.g. paragraph alignment, and therefore can get back on track after an alignment error (or in spite of a translation discrep
ancy).
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {To take an example, we could imagine the following chunks of text to be aligned:
\par 
\par }\trowd \trgaph70\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx4111\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8505\pard\plain \s33\qj\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright 
\fs20\cgrid {English original (extract from Doris Lessing's }{\i The Good Terrorist}{)\cell Portuguese translation by Bernardette Pinto Leite\cell }\pard\plain \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\fs20 
\row }\trowd \trgaph70\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx4111\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8505\pard\plain \s33\qj\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright 
\fs20\cgrid {<s>She faced him, undefiant but confident, and said, "I wonder if they will accept us?"</s> <s>And, as she had known he would, he said, "It is a question of whether we will accept them."</s></p>
\par <p><s>She had withstood the test on her, that bony pain, and he let her wrist go and went on to the door.</s>\cell <p><s>Ela encarou-o, sem desafio mas confiante, e perguntou:</s></p>
\par <p><s>&mdash; Achas que nos aceitam?</s> </p>
\par <p><s>&mdash; E, conforme sabia que Jasper responderia, este retorquiu:</s></p>
\par <p><s>&mdash; \'c9 tudo uma quest\'e3o de n\'f3s os aceitarmos a eles.</s></p>
\par <p><s>Alice resistira ao teste sobre a sua pessoa, \'e0 dor \'f3ssea, e ele largou-lhe o pulso e dirigiu-se para a porta.</s>\cell }\pard\plain \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\row }\pard\plain 
\s21\qc\sa120\sl-220\slmult0\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \i\fs20\cgrid {Figure 1. }{\i0 Texts to be aligned}{
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {We would expect that some words in the English extract would match some of the Portuguese words i
n the anchor word list; these matches would in turn enhance the possibility of the program linking the correct sentences. For the second English sentence in the extract above, for instance, we will get the following matches in the word list:}{
\cs26\cf6\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ This is the original format o
f the TCA wordlist. "*" means truncation, "/ "separates source and target items, "," separates items inside each language.}}}{
\par }\pard\plain \s33\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {
\par and e 
\par is, 's / \'e9, est\'e1
\par question* / pergunt*, quest*
\par we n\'f3s
\par accept* aceit*
\par them / lhes, os, as
\par 
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {
The TCA assigns a unique identification to each s-unit with its corresponding s-unit(s) in the translation. When the original sentence has only one corresponding sentence i
n the translation, we get a 1:1 correspondence. When the original sentence has been translated into two sentences, we get a 1:2 correspondence:
\par 
\par }\pard\plain \s33\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\f6\fs18 <s id=DL2.1.s16 corresp='DL2TP.1.s17 DL2TP.1.s18'>And, as she had known he would, he said, "It is a question of whether we will accept them."</s>

\par 
\par <s id=DL2TP.1.s17 corresp=DL2.1.s16>&mdash; E, conforme sabia que Jasper responderia, este retorquiu:</s>
\par <s id=DL2TP.1.s18 corresp=DL2.1.s16>&mdash; \'c9 tudo uma quest\'e3o de n\'f3s os aceitarmos a eles.</s>
\par 
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {Each text has a u
nique code, normally starting with the authors' initials, e.g. a text by Doris Lessing has a code DLx. All texts will be referred to by their code.
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 3.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid {Human intervention required
\par }\pard\plain \s34\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {The program handles 1:1, 1:2, and 1:0 correspondences, i.e. one s-unit matching 
one, two or zero s-units in the translation; the latter two have to be checked manually. The remaining matrices, containing 1:1 correspondences only, are assumed to be correct and are not systematically checked. Hence, the file that is proofread only incl
udes matrices that do not contain 1:1 correspondences throughout.
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {
Table 1 shows the percentage of matrices that the proofreader has to check, an average of 48.8%, for the 16 English-Portuguese texts. This apparently discouraging percentage needs some explan
ation since it does not reflect the actual manual intervention that takes place. The picture becomes skewed simply because we immediately associate half of the matrices with half of the sentences in a text. This is not the case, however. Each matrix conta
i
ns approximately 10-12 sentence pairs, and as will be shown in the matrix below (Figure 2), the proofreader will only have to investigate the s-units that are not 1:1 correspondences. This is to say that although the proofreader has to investigate almost 
half of the matrices, he will not have to investigate half of the text/s-units, but merely a small percentage.
\par }\pard\plain \s19\qc\li113\ri113\sb120\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\cs18\i Table 1}{. Number of matrices to check
\par }\trowd \trqc\trrh1134\trleft-106\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx958\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1897\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2771\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3661\pard\plain \qc\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0
\pndec }\adjustright \lang2057\cgrid {\fs20 Text\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 Matrices in output file}{\cs26\fs20\super \chftn {\footnote\ftnalt \pard\plain 
\s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ In other words, the total number of matrices in the text.}}}{\fs20 \cell Matrices to check\cell %\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\trowd \trqc\trgaph108\trleft-214\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv
\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx850\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb 
\cellx1789\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2663\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3553
\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 ABR1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 120\cell 78\cell 65\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 AH1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 131\cell 98\cell 74\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 AT1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 113\cell 30\cell 27\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 BC1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 94\cell 39\cell 41\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 DL1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 90\cell 54\cell 60\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 DL2\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 139\cell 100\cell 72\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 FF1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 88\cell 76\cell 86\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JB1P\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 97\cell 32\cell 33\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JB1PP\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 97\cell 32\cell 33\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JH1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 61\cell 20\cell 33\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 MA1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 74\cell 6\cell 8\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 NG1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 74\cell 37\cell 50\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 PDJ3\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 108\cell 49\cell 45\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 RDO1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 143\cell 52\cell 36\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 ST1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 128\cell 87\cell 68\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 WB1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 87\cell 12\cell 14\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trgaph108\trleft-214\trkeep\trbrdrt
\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx850
\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1789\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2663\clvertalt
\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3553\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 Total\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 1,643\cell 802\cell 48.8\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \row }\pard 
\li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {To take an example, three of the sentence pairs in the 
matrix in Figure 2 have to be manually checked, and if necessary, corrected. This can be seen from the low values given in the diagonal of the matrix below (top left to bottom right). Moreover, the numbers below the matrix reflect this; the English senten
c
e 4, for instance, is said to correspond to 0, sentence 5 to Portuguese sentences 4+5, etc. It can be seen, then, that sentence 4 in the English original is not linked to any sentence in the translation, and wrongly so. We will therefore have to match it 
to the translation \endash  sentence 3 in Portuguese \endash  manually, correcting the alignment to a 1:2 correspondence. The other two 1:2 correspondences found in the matrix did not have to be corrected.
\par }\pard\plain \li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {
\par }\pard \fi720\li720\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\f6\fs20\ul 84  20  71  51  23  56  48 106 133  51 
\par }\pard \li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\f6\fs20              }{\f6\fs20\ul 1   2   3   4   5   6   7   8   9  10 }{\f6\fs20 \tab Port. trans.
\par           ------------------------------------------------
\par   }{\i\f6\fs20 1  97}{\f6\fs20  I    }{\b\f6\fs20 6 }{\f6\fs20   0   1   2   0   0   0   2   1   1 
\par   }{\i\f6\fs20 2  18}{\f6\fs20  I    1   }{\b\f6\fs20 2}{\f6\fs20    0   0   0   0   0   1   0   0 
\par   }{\i\f6\fs20 3  58}{\f6\fs20  I    2   0   }{\b\f6\fs20 0}{\f6\fs20    1   0   1   0   2   2   0 
\par   }{\i\f6\fs20 4   7}{\f6\fs20  I    0   0   0   }{\b\f6\fs20 0}{\f6\fs20    0   0   0   0   0   0 
\par   }{\i\f6\fs20 5  78}{\f6\fs20  I    2   0   1   }{\b\f6\fs20 4}{\f6\fs20    }{\b\f6\fs20 3}{\f6\fs20    1   2   3   1   0 
\par   }{\i\f6\fs20 6  85}{\f6\fs20  I    3   1   1   4   2   }{\b\f6\fs20 2}{\f6\fs20    }{\b\f6\fs20 6}{\f6\fs20    3   1   3 
\par   }{\i\f6\fs20 7  96}{\f6\fs20  I    2   2   1   3   1   2   1   }{\b\f6\fs20 7}{\f6\fs20    2   1 
\par   }{\i\f6\fs20 8 124}{\f6\fs20  I    1   0   0   1   0   1   0   4   }{\b\f6\fs20 6}{\f6\fs20    0 
\par }{\i\f6\fs20   9  51}{\f6\fs20  I    1   0   1   2   1   2   2   1   1   }{\b\f6\fs20 2}{\f6\fs20  
\par  }{\i\f6\fs20 10 164}{\f6\fs20  I    1   1   1   4   2   2   2   6   2   }{\b\f6\fs20 0}{\f6\fs20  
\par }\pard\plain \s20\qj\li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\f6  \u8593\'3f
\par }\pard\plain \li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\f6\fs20 Eng. orig.
\par }\pard \fi720\li720\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\f6\fs18 1,1 2,2 3,3 4,0 5,4+5 6,6+7 7,8 8,9 9,10
\par }\pard \li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\f6\fs18 ---------------------------------------------------------------------------
\par  1: <s>For her part she did not have to be told that she was wearing <hi 
\par      rend=italic>her look</hi>, described by him as silly.</s>  (DL2.1.11)
\par 
\par  1: <s>Quanto a ela, sabia que estava com o <hi rend=italic>seu olhar</hi>, 
\par      que ele descrevia como de aparvalhado.</s></p>  (DL2TP.1.12)
\par ---------------------------------------------------------------------------
\par  2: <s>"Stop it," he ordered.</s>  (DL2.1.12)
\par 
\par  2: <p><s>&mdash; P\'e1ra &mdash; ordenou ele.</s>  (DL2TP.1.13)
\par ---------------------------------------------------------------------------
\par  3: <s>His hand shot out, and her wrist was encircled by hard bone.</s>  (DL2.1.13)
\par 
\par  3: <s>Estendendo a m\'e3o, apertou com for\'e7a o pulso da rapariga, causando-lhe dor.</s></p>  (DL2TP.1.14)
\par ---------------------------------------------------------------------------
\par  4: <s>It hurt.</s>  (DL2.1.14)
\par 
\par 
\par ---------------------------------------------------------------------------
\par  5: <s>She faced him, undefiant but confident, and said, "I wonder if they will accept us?"</s>  (DL2.1.15)
\par 
\par  4: <p><s>Ela encarou-o, sem desafio mas confiante, e perguntou:</s></p>  (DL2TP.1.15)
\par  5: <p><s>&mdash; Achas que nos aceitam?</s></p>  (DL2TP.1.16) 
\par ---------------------------------------------------------------------------
\par  6: <s>And, as she had known he would, he said, "It is a question of whether we will accept them."</s></p>  (DL2.1.16)
\par 
\par  6: <p><s>&mdash; E, conforme sabia que Jasper responderia, este retorquiu:</s></p>  (DL2TP.1.17)
\par  7: <p><s>&mdash; \'c9 tudo uma quest\'e3o de n\'f3s os aceitarmos a eles.</s></p>  (DL2TP.1.18)
\par ---------------------------------------------------------------------------
\par  7: <p><s>She had withstood the test on her, that bony pain, and he let her wrist go and went on to the door.</s>  (DL2.1.17)
\par 
\par  8: <p><s>Alice resistira ao teste sobre a sua pessoa, \'e0 dor \'f3ssea, e ele largou-lhe o pulso e dirigiu-se para a porta.</s>  (DL2TP.1.19)
\par ---------------------------------------------------------------------------
\par  8: <s>It was a front door, solid and sure of itself, in a little side street full of suburban gardens and similar comfortable houses.</s>  (DL2.1.18)
\par 
\par  9: <s>Era uma porta de entrada s\'f3lida, segura, situada numa ruazinha secund\'e1ria com jardins de sub\'farbios e casas semelhantemente confort\'e1veis.</s>  (DL2TP.1.20)
\par ---------------------------------------------------------------------------
\par  9: <s>They did not have slates missing and broken windows.</s></p>  (DL2.1.19)
\par 
\par }\pard\plain \s15\li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \f6\fs20\lang2057\cgrid {\fs18 10: <s>N\'e3o lhes faltavam telhas nem tinham vidros partidos.</s></p>  (DL2TP.1.21)
\par }\pard\plain \li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\f6\fs18 ---------------------------------------------------------------------------
\par }\pard\plain \s21\qc\li113\ri113\sa120\sl-220\slmult0\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \i\fs20\cgrid {Figure 2. }{\i0 Example of a matrix calculated by the Translation Corpus Aligner}{\cs26\i0\fs16\super \chftn 
{\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ The first line of underscored numbers gives the number of characters in
 each Portuguese sentence, the second line the sentence number. The vertical italicized lines give the sentence number of each English sentence and its length in number of characters.}}}{
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {After the percentage of matrices to be checked had been calculated, the next step was to find out how many corrections one actually had to m
ake. Table 2 gives the percentages of corrections that were made after alignment.
\par Table 2 presents, for each text, the number of s-units that need to be corrected after alignment with the anchor word list. Since not all s-units have actually been inspected
 by the proofreader (the percentage of matrices was shown in Table 1), the two rightmost columns give the number of s-units inspected manually, and the corresponding correction percentage.
\par }\pard\plain \s19\qc\li113\ri113\sb120\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\i Table 2.}{ Number of corrected s-units after running the program with the anchor list
\par }\trowd \trqc\trrh1134\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx959\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1809\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx2801\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3794\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx4786\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5636\pard\plain \ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright 
\lang2057\cgrid {\fs20 Text\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 Corrections\cell Total number of  s-units\cell Corrected s-units (% of  total)\cell Number of s-units in the matrices inspected}{
\cs26\fs20\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ We assume that the percentage of s-units inspected is equal to the per
centage of matrices shown, therefore computing the present numbers by multiplying the total number of s-units by the percentage of matrices to check (displayed in Table 1). This is an approximation, because the same s-unit can belong to two contiguous mat
rices.}}}{\fs20 \cell Corrected s-units (%  of inspected)\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr
\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx959\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1809\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2801\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3794\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4786\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5636\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 ABR1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\fs20 33\cell 1,139\cell 2.9\cell 740\cell 4.4\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 AH1
\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 42\cell 1,263\cell 3.3\cell 934\cell 4.5\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 AT1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 19\cell 1,102\cell 1.7\cell 297\cell 6.4\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 BC1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 25\cell 893\cell 2.8\cell 366\cell 6.8\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 DL1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 11\cell 855\cell 1.3\cell 513\cell 2.1\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 DL2\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 61\cell 1,307\cell 4.7\cell 941\cell 6.5\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 FF1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 48\cell 713\cell 6.7\cell 613\cell 7.8\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JB1P\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 12\cell 934\cell 1.3\cell 308\cell 3.9\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JB1PP\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 23\cell 927\cell 2.5\cell 305\cell 7.5\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JH1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 15\cell 584\cell 2.6\cell 192\cell 7.8\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 MA1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 2\cell 730\cell 0.3\cell 58\cell 3.4\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 NG1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 12\cell 702\cell 1.7\cell 351\cell 3.4\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 PDJ3\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 14\cell 1,041\cell 1.3\cell 468\cell 3.0\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 RDO1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 42\cell 1,396\cell 3.0\cell 502\cell 8.4\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 ST1\cell }\pard \qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 50\cell 1,204\cell 4.2\cell 818\cell 6.1\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 WB1\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 5\cell 727\cell 0.7\cell 101\cell 5.0\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trkeep\trbrdrt
\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx959
\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1809\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2801\clvertalt
\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3794\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4786\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5636\pard \li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 Average\cell }\pard 
\qc\li113\ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 \cell \cell 2.9\cell \cell 5.5\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\li113\ri113\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {\lang2057 
We see that the proofreader has to make changes in about 5.5% of the s-units inspected, which corresponds to about 2.9% of all s-units present in the corpus. Again it should
 be pointed out that the number of s-units found in the matrices that are inspected is considerably lower than the number of s-units actually inspected.
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 4.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid {The importance of the anchor word list
\par }\pard\plain \s34\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {We now proceed to evaluate the importance of the anchor word lis
t, by comparing the amount of revision and modification required when running the alignment program with and without language specific information (that is, with an empty anchor list). In order to estimate the differences, we ran the program with and with
out the anchor word list and then compared automatically the differences with the final (post-edited) version.
\par }\pard\plain \s19\qc\li113\ri113\sb120\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\i Table 3.}{ Number of differences in the alignment of English-Portuguese texts 
\par }\trowd \trqc\trrh1134\trleft108\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1101\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1952\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2944\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4078\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5070\pard\plain \ri113\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\fs20 Text\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\fs20 No. of "skips"}{\cs26\fs20\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{
 It is important to note that in the process described so far, so called skip-attributes had already been inserted in the text files. Skip attributes are commands for the program to "skip" a particular sentence, and they are one of the few tags and 
attributes added by the ENPC project to the basic recomendations put forward by the Text Encoding Initative (Sperberg-McQueen & Burnard, 1994). The purpose of skip-attributes (in either language) is to mark that a sentence has not been translated, or has 
been invented. Although this is a sort of human intervention, it is hard to quantify, because it is usually inserted in an interactive mode when and if the program does not manage to continue. Here we have chosen to disregard them. }}}{\fs20 \cell 
No. of differences (final version vs. raw version with anchor word list)\cell No. of differences (final version vs. raw version w/o anchor word list)\cell No. of differences
\par (with vs. w/o anchor word list)\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trgaph108\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh
\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx993\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx1844\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2836\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx3970\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4962\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 ABR1\cell 
}\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 8\cell 33\cell 139\cell 126\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 AH1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 3\cell 44\cell 111\cell 88\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 AT1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright 
{\fs20 4\cell 27\cell 38\cell 39\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 BC1*\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 2\cell 38\cell \cell \cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0
\pndec }\adjustright {\fs20 DL1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 2\cell 14\cell 31\cell 30\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 DL2*\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 6\cell 64\cell \cell \cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 FF1*\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\fs20 22\cell 60\cell \cell \cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JB1P*\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 3\cell 18\cell \cell \cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0
\pndec }\adjustright {\fs20 JB1PP\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 5\cell 31\cell 45\cell 50\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 JH1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 2\cell 25\cell 19\cell 35\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 MA1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright 
{\fs20 0\cell 4\cell 12\cell 14\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 NG1*\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 0\cell 15\cell \cell \cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0
\pndec }\adjustright {\fs20 PDJ3\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 2\cell 19\cell 11\cell 26\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 RDO1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 2\cell 67\cell 72\cell 65\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 ST1*\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\fs20 3\cell 68\cell \cell \cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trgaph108\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh
\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx993\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx1844\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2836\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx3970\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4962\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 WB1\cell 
}\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\fs20 8\cell 6\cell 14\cell 9\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0
\pndec }\adjustright {
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {The results are shown in Table 3. The texts marked with a star (six out of sixteen)}{\cf6  }{could not be aligned without the 
anchor word list. Further, Table 3 shows the differences between the final, proofread versions of the ENPC texts and the versions with and without the anchor word list prior to revision.}{\cs26\super \chftn {\footnote\ftnalt \pard\plain 
\s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ These differences were computed automatically with the help of a simple Perl program (}{\f6 compara_alinhamento.pl}{
), running after the Unix command }{\f6 diff}{\i  }{was invoked between the resulting files.}}}{ 
\par Due to the simplicity of the (programming) approach (only comparing the target part), missing s-units in the Portuguese translations, such as the (rare) example in Figure 3, were not identified:
\par }\pard\plain \s15\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \f6\fs20\lang2057\cgrid {\fs18 
\par 6: <s>If we got married, we could no longer go back, whether we wanted to or not.</s>  (ABR1.1.1.242)
\par }\pard\plain \s20\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\f6\fs18\lang2057 
\par }\pard\plain \widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\f6\fs18  6: <s>Se nos cas\'e1ssemos, n\'e3o poder\'edamos mais voltar, quer quis\'e9ssemos ou n\'e3o.</s>  (ABR1TP.1.247)
\par ---------------------------------------------------------------------------
\par  7: <s>Neither he nor I.</s>  (ABR1.1.1.243)
\par 
\par ---------------------------------------------------------------------------
\par  8: <s>He white; I coloured.</s>  (ABR1.1.1.244)
\par }\pard\plain \s23\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\f6\fs18 
\par }\pard\plain \widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {\f6\fs18  7: <s>Ele era branco, eu mesti\'e7a.</s>  (ABR1TP.1.248)
\par ---------------------------------------------------------------------------}{\f6\fs20 
\par }\pard\plain \s21\qc\sa120\sl-220\slmult0\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \i\fs20\cgrid {Figure 3. }{\i0 Example of missing Portuguese translation, overlooked by the }{\i0\f6\fs18 compara_alinhamento.pl}{\i0  program}{  

\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {
It is clear that the English-Portuguese anchor word list does help the alignment program and reduces the number of changes to be made by the proofreader, though perhaps not as markedly as one might expect. Howe
ver, the fact that 6 out of the 16 texts did not make it through alignment indicates that the program to a large extent depends on the anchor word list, and not only on the other factors mentioned in Section 2.
\par By using the anchor word list, the percentage of sentences to be corrected drops from 4 to 2% of all the sentences that are manually inspected (which, in turn, can be considered approximately one eighth of all sentences).}{\cs26\super \chftn 
{\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ Assuming that for each matrix to be observed, 3-4 sentences have to be read by the human
 reviewer, this means a fourth of all sentences present in the matrices. Then, only half the matrices are revised: \'bd*1/4=1/8 }}}{ Again, let us stress that the anchor word list was not originally made for the
 language pair English-Portuguese. Neither of the anchor word lists was corpus driven; the original anchor list, for English-Norwegian, was manually encoded based on the intuition of the linguists working on the ENPC previous to the choice of the actual t
exts.}{\cs26\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{
 Knut Hofland (p.c.) revised and improved the list, using a program that computed a threshold of usefulness for very frequent items, and removing those which would not help the alignment.}}}{ 
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 5.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid { Considering the anchor word list in detail
\par }\pard\plain \s34\qj\widctlpar\adjustright \cgrid {
Would more attention to the English-Portuguese pair pay off? What would one do in order to optimize the performance for this language pair, and what would be the net gain? This is what we set out to test in the next step.
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {We have thus created a program which, for each English source text and its corresponding Portuguese translation, counts 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls26\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls26\adjustright \cgrid {
the number and percentage of the English anchor entries in the English text
\par {\pntext\pard\plain\f3\lang2057\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \fi-360\li1788\widctlpar\jclisttab\tx1788{\*\pn \pnlvlblt\ilvl0\ls2\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls2\adjustright \lang2057\cgrid {
compared to the entries in the anchor list
\par {\pntext\pard\plain\f3\lang2057\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \fi-360\li1788\widctlpar\jclisttab\tx1788{\*\pn \pnlvlblt\ilvl0\ls2\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls2\adjustright {
in terms of the s-units the matches refer to
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
the number and percentage of the Portuguese anchor entries in the Portuguese text
\par {\pntext\pard\plain\f3\lang2057\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \fi-360\li1788\widctlpar\jclisttab\tx1788{\*\pn \pnlvlblt\ilvl0\ls2\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls2\adjustright \lang2057\cgrid {
compared to the entries in the anchor list
\par {\pntext\pard\plain\f3\lang2057\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \fi-360\li1788\widctlpar\jclisttab\tx1788{\*\pn \pnlvlblt\ilvl0\ls2\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls2\adjustright {
compared to the total number of target s-units 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
the actual successful matches (i.e., the cases where both members of the anchor pair occur in a translation pair)
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
the ratio of successful matches vs. all possible matches
\par {\pntext\pard\plain\f3\lang2057\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \fi-360\li1788\widctlpar\jclisttab\tx1788{\*\pn \pnlvlblt\ilvl0\ls2\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls2\adjustright \lang2057\cgrid {
in terms of the occurrences of the source member of the anchor list
\par {\pntext\pard\plain\f3\lang2057\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \fi-360\li1788\widctlpar\jclisttab\tx1788{\*\pn \pnlvlblt\ilvl0\ls2\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls2\adjustright {
in terms of the occurrences of the target member of the anchor list
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {In order to compute these numbers, several decisions have to be made: 
\par First of all, the anchor word list is unwrapped from the source side, i.e., the cases of A,B / C are transformed into A / C and B / C, resulting in 1,022 pairs (from an or
iginal anchor list containing 882 lines). On the other hand, this is not done for the Portuguese side, since it is understood that E / F, G would succeed in either case. In Figure 4, we provide some illustration of what the unwrapping does:}{\cs26\super 
\chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ The "." and the parentheses in the reformulation are simply the rendition of the same semantics in Perl}{\lang2070 
 syntax and have no significance for the success of the pattern matching.}}}{
\par 
\par }\trowd \trqc\trgaph108\trleft-380\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 
\clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3026\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6286\pard\plain \s15\widctlpar\intbl\adjustright \f6\fs20\lang2057\cgrid {\f0\fs18 is, 's / 
\'e9, est\'e1 \cell 's / ((\'e9)|(est\'e1))
\par is / ((\'e9)|(est\'e1))\cell }\pard\plain \widctlpar\intbl\adjustright \lang2057\cgrid {\fs18 \row }\trowd \trqc\trgaph108\trleft-380\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 
\trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3026\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 
\cltxlrtb \cellx6286\pard\plain \s15\widctlpar\intbl\adjustright \f6\fs20\lang2057\cgrid {\f0\fs18 English* / ingl* \cell English.* / ingl.*\cell }\pard\plain \widctlpar\intbl\adjustright \lang2057\cgrid {\fs18 \row }\pard\plain 
\s15\widctlpar\intbl\adjustright \f6\fs20\lang2057\cgrid {\f0\fs18 became, becom* / torn*, volt*, fic*\cell became / ((torn.*)|(volt.*)|(fic.*))
\par becom.* / ((torn.*)|(volt.*)|(fic.*))\cell }\pard\plain \widctlpar\intbl\adjustright \lang2057\cgrid {\fs18 \row }\pard\plain \s15\widctlpar\intbl\adjustright \f6\fs20\lang2057\cgrid {\f0\fs18 has, have, 've / tenho, tens, tem, temos, t\'eam
\par \cell has  / ((tenho)|(tens)|(tem)|(temos)|(t\'eam))
\par have / ((tenho)|(tens)|(tem)|(temos)|(t\'eam))
\par 've / ((tenho)|(tens)|(tem)|(temos)|(t\'eam))}{\cs26\f0\fs18\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ From the unwrapp
ing we see clearly that, given that Portuguese has more verbal inflections than English (whereas the opposite happens to Norwegian), one should have written the pairs }{\f6\fs18 has/tem}{ and }{\f6\fs18 have, 've/tenho, tens, temos, t\'eam}{ instead.}}}{
\f0\fs18 \cell }\pard\plain \widctlpar\intbl\adjustright \lang2057\cgrid {\fs18 \row }\trowd \trqc\trgaph108\trleft-380\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv
\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3026\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb 
\cellx6286\pard\plain \s15\widctlpar\intbl\adjustright \f6\fs20\lang2057\cgrid {\f0\fs18 7*, seven / 7*, sete\cell 7.* : ((7.*)|(sete))
\par seven : ((7.*)|(sete))\cell }\pard\plain \widctlpar\intbl\adjustright \lang2057\cgrid {\fs18 \row }\pard\plain \s21\qc\sa120\sl-220\slmult0\widctlpar\adjustright \i\fs20\cgrid {Figure 4. }{\i0 Original and modified anchor list}{ 
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {Then, pattern matching is done so that the matches are only counted in word contexts. Parts of words were not counted as successful matches (so }{\i are}{ does not succeed in }{\i mare}{
, for example), except when the pattern expression explicitly says so (like in }{\i becom*}{).
\par In addition, the matching ignores case. This is justified due to the low probability of occurrence of non-capitalized instances of words that require capitalization (such as *}{\i english}{ or *}{\i inglaterra}{).
\par The most important decision, as well as the one which may be less obvious, is tha
t we only count 1:1 matches as being successful. That is, "successful pairs" in Table 4 below include only those in which the source member is found in a source s-unit which has }{\b one}{
 single sentence corresponding to it and in which, moreover, (one of) the target member(s) of the anchor word list is found. The reason for this restrictive computation of anchor list successes is twofold: If mappings of 1:2 or 1:3 were found, 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li360\widctlpar\jclisttab\tx360{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
they would have been subjected to the revisor's consideration anyway 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li360\widctlpar\jclisttab\tx360{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {it would be cons
iderably more difficult to quantify both the success (should one use a pondered average of target s-units instead of natural numbers?) and the usefulness of the information in the anchor word list}{\cs26\super \chftn {\footnote\ftnalt \pard\plain 
\s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ For example, one would expect th
e presence of the target member of the pair in a sentence contiguous to the right one (the one in the final alignment) as a measure of the negative influence a particular word-pair would have. When one sentence is aligned with two sentences, one of which 
w
ould quantify as success in terms of our computations, one could expect a favourable import from the anchor word list if the alignment is right, and the opposite if it is not right. So, a good measure of the anchor list doings would have to take into acco
unt the percentage of e.g. 1:2 alignments that were correct. To do this would most probably require mimicking the way the program works, which is outside the scope of the present paper.}}}{
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {Finally, it should be noted that the anchor list included the entry }{\i &mdash}{ / }{\i &mdash}{
, which might not look specific to the English-Portuguese pair. However, some punctuation \endash  and its translation \endash  is language specific (Santos, 1998b), and therefore the anchor pair including }{\i &mdash}{ was kept in the list for evaluat
ion.}{\cs26\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{ In fact, even "invariants" such as numbers can actually result in a d
ifferent translation. Floors are counted differently in different languages: ground floor in Portuguese is Norwegian first floor; soccer's first division in Norway is second division in Portugal; a person measured in feet will not have a corresponding hei
ght in meters, etc. And Figure 4 reminds us that }{\i seven hundred}{ does not get translated into }{\i sete centos}{ but into }{\i setecentos}{ (as opposed to Norwegian, which has two words as well). }}}{
\par }\pard\plain \s2\sb240\sa60\keepn\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\outlinelevel1\adjustright \b\f4\fs28\lang2057\cgrid {5.1 Looking at the occurrences of anchor items in the corpus
\par }\pard\plain \s34\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {
Table 4 presents the quantitative results for each pair of texts. As far as the English matches are concerned, the first column displays how many members of the anchor list are actually pre
sent in the source text (the percentage relative to the total number of anchor pairs is given in the second column). The number of occurrences in the text of the source expressions of the anchor word list is given in the third column under "English matche
s
". The number of different (target) s-units corresponding to some match of the English expressions in the source text is shown in the fourth column (i.e., more than one match of the same English expression per target s-unit is disregarded). As far as Port
u
guese matches are concerned, the corresponding information is provided: The next two columns give the number and percentage of target expressions of the anchor list which were found in the target texts, together with the number of actual matches. The most
 
interesting information is found under the heading "Successful pairs", which displays how many times pairs in the anchor list were actually found in a translation pair, and what the percentage is relative to the simple occurrence of one element of the pai
r.
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {
The success matches correspond roughly to a fourth of the source units in which they are found and to less than a tenth of the target units in which they appear. I.e., the relevance of the target expressions seems to be considerably lower than that of th
e source expressions. This can be explained by several distinct factors. First of all, the English terms were originally chosen with a view to finding a good translation correspondence with respect to Norwegian. This is not necessarily the case for the En
glish-Portuguese pair. Then, the linguist adapting the list}{\cs26\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\lang2057\cgrid {\cs26\super \chftn }{
 Incidentally, one of the authors of the present paper (D.S.).}}}{ failed to note that some target expressions were too general (matching, for example, unrelated and very frequent Portuguese prepositions \endash 
 example in (4) below), which obviously diminishes the per
centage of relevant pairs. Note that this does not necessarily diminish the program's performance, since there is no reason to suppose that the TCA looks for every occurrence in the target language, as we did here for evaluation purposes.
\par }\pard\plain \s19\qc\sb120\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\i \page Table 4.}{ Coverage of the anchor list for the English-Portuguese corpus
\par }\trowd \trqc\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb 
\cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3402\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5529
\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8222\pard\plain \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {Text\cell English matches
\cell Portuguese matches\cell Successful pairs\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 
\trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 
\cltxlrtb \cellx1985\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3402\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb 
\cellx4678\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5529\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6426
\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7324\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8222\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\cell anchor list\cell s-units\cell anchor list\cell s-units\cell number\cell source\cell target\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\trowd \trqc\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1418\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1985\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2694\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3402\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4110\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4678\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5528\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6426\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7324\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8222\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {ABR1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {695\cell 68%\cell 7,194
\cell 6,454\cell 780\cell 76%\cell 50,226\cell 3,585\cell 49.8%\cell 7.1%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {AH1\cell 
}\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {655\cell 64%\cell 6,865\cell 6,206\cell 723\cell 71%\cell 51,147\cell 2,981\cell 43.4%\cell 5.8%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {AT1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {624\cell 61%\cell 5,988\cell 5,361\cell 739\cell 72%\cell 48,585
\cell 3,177\cell 53.1%\cell 6.5%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {BC1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {700\cell 68%\cell 4,646\cell 3,933\cell 795\cell 78%\cell 38,381\cell 2,500\cell 53.8%\cell 6.5%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {DL1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {609\cell 60%\cell 6,400\cell 5,783\cell 716\cell 70%\cell 45,860
\cell 3,328\cell 52.0%\cell 7.3%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {DL2\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {623\cell 61%\cell 7,014\cell 6,375\cell 739\cell 72%\cell 48,279\cell 2,761\cell 39.4%\cell 5.7%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {FF1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {690\cell 68%\cell 6,063\cell 5,246\cell 776\cell 76%\cell 39,095
\cell 2,505\cell 41.3%\cell 6.4%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {JB1P\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {616\cell 60%\cell 6,056\cell 5,434\cell 737\cell 72%\cell 47,353\cell 3,507\cell 57.9%\cell 7.4%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {JB1PP\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {616\cell 60%\cell 6,063\cell 5,437\cell 758\cell 74%\cell 
42,989\cell 2,866\cell 47.3%\cell 6.7%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {JH1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {623\cell 61%\cell 5,040\cell 4,417\cell 743\cell 73%\cell 39,991\cell 2,996\cell 59.4%\cell 7.5%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {MA1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {615\cell 60%\cell 5,448\cell 4,833\cell 735\cell 72%\cell 39,231
\cell 3,011\cell 55.2%\cell 7.7%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {NG1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {675\cell 66%\cell 6,536\cell 5,857\cell 746\cell 73%\cell 43,566\cell 3,717\cell 56.9%\cell 8.5%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {PDJ3\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {718\cell 70%\cell 7,632\cell 6,910\cell 803\cell 79%\cell 
55,941\cell 4,072\cell 53.4%\cell 7.3%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {RDO1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {538\cell 53%\cell 6,522\cell 5,980\cell 665\cell 65%\cell 43,813\cell 3,880\cell 59.5%\cell 8.8%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {ST1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {692\cell 68%\cell 6,024\cell 5,318\cell 786\cell 77%\cell 49,826
\cell 2,781\cell 46.1%\cell 5.6%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {WB1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {629\cell 62%\cell 4,662\cell 4,030\cell 728\cell 71%\cell 32,463\cell 2,675\cell 57.4%\cell 8.2%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }
\adjustright {\row }\trowd \trqc\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1418\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1985\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2694\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3402\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4110\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4678\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5528\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6426\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7324\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb
\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8222\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {Average\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {645\cell 63%\cell 
\cell \cell 748\cell 73%\cell \cell \cell 51.6%\cell 6.6%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {In any case, we find a significant number of matches, even if several pairs (at least around a fourth) do not appear in the files. If we divide t
he number of successes by the number of pairs whose source expression is present in the source text, we get on average 5 matches per pair. Later on we will look more closely at the significance of some elements of the anchor list.
\par Still, the number of succ
essful pairs in Table 4 does not really offer a good illustration of the contribution of the anchor list, since they take each success independently of the place it occurs in. For example, for a particular s-unit there might be 10 successful matches with 
ten different pairs, while for another not a single one. This means that the values in the last two columns do }{\ul not}{
 express the percentages of s-units that received positive match in the anchor list, but simply the percentage of occurrence of (source or target) expressions that have been useful for alignment.
\par So, we made our program also count the (target) s-units for which there was }{\ul one or more}{ matches for any anchor pair (i.e., the s-units which corresponded to success in finding both the source expression 
and the target expression). The program also provided the percentage of s-units in terms of the number of translation pairs in the file. These results, which are much easier to interpret, are displayed in Table 5: The first column shows the number of matc
h
ing anchor-list pairs (after restructuring), and the percentage is given in column 2. The third column shows the number of s-units which had one or more successful matches in the anchor list, and column 4 lists the total number of translation pairs in eac
h file (counted in the target file) for ease of reference. Column 5 is simply the result of dividing the value of column 3 by that of column 4, giving the percentage of translation pairs with successful matches in the anchor list.
\par We see from this table that the great majority of sentences have hits in the anchor pairs, which may indicate that, when a pair actually occurs in the text, it gives some positive result for the alignment process.
\par }\pard\plain \s19\qc\sb120\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \fs20\cgrid {\i Table 5.}{ Coverage of the successful anchor pairs in terms of anchor list and s-units
\par }\trowd \trqc\trleft3\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb 
\cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4149\pard\plain \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \lang2057\cgrid {Text\cell Successes
\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trleft2\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv
\brdrs\brdrw10 \clvertalt\clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1985\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2671\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3379\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4148\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\cell anchor list\cell s-units\cell total\cell %\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trleft2\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt
\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1418\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1985\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2671\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3379\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4148\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {ABR1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {563\cell 55%\cell 1014\cell 1139\cell 89%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {AH1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {488
\cell 48%\cell 1069\cell 1263\cell 85%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {AT1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {496\cell 49%\cell 1030\cell 1102\cell 93%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {BC1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {557\cell 55%\cell 823\cell 893\cell 92%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {DL1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {492
\cell 48%\cell 783\cell 855\cell 92%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {DL2\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {482\cell 47%\cell 1120\cell 1307\cell 86%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {FF1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {512\cell 50%\cell 706\cell 713\cell 99%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {JB1P\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {479
\cell 47%\cell 882\cell 934\cell 94%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {JB1PP\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {464\cell 45%\cell 879\cell 927\cell 95%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {JH1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {490\cell 48%\cell 572\cell 584\cell 98%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {MA1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {508
\cell 50%\cell 711\cell 730\cell 97%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {NG1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {526\cell 51%\cell 653\cell 702\cell 93%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {PDJ3\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {586\cell 57%\cell 970\cell 1041\cell 93%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {RDO1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {430
\cell 42%\cell 1221\cell 1396\cell 87%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {ST1\cell }\pard 
\qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {521\cell 77%\cell 870\cell 1204\cell 72%\cell }\pard \widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {WB1\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {515\cell 50%\cell 719\cell 853\cell 84%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\trowd \trqc\trleft2\trkeep\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt
\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx851\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1418\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1985\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2671\clvertalt\clbrdrt
\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3379\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4148\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {Average\cell }\pard \qc\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {506\cell 50%\cell \cell \cell 91%\cell }\pard 
\widctlpar\intbl{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright {\row }\pard\plain \s2\sb240\sa60\keepn\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\outlinelevel1\adjustright \b\f4\fs28\lang2057\cgrid {5.2 Look
ing at some anchor pairs in particular
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {
We expect, however, that not all anchor pairs have the same import, and in fact we see from Table 5 that, on average, only half of them work for each text. That is why we wanted to take a closer look at the results.
\par From a rough inspection of the results, and along with more accurate pairs, the two following cases were easy to detect:
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
systematic translations were forgotten (like the weapon sense of }{\i arm}{* or the}{\i  artist* }{continuation of}{\i  art*}{)
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li1068\widctlpar\jclisttab\tx1068{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
translation involves too much restructuring, due to language differences, so that lexical clues are not the right place to look (as is the case of }{\i I}{ or }{\i be}{).
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {In order to look more closely at the results, we selected several frequent cases \endash  pairs that were most successful (in absolute terms) \endash 
 and looked at their distribution in the different texts. Table 6 shows the number of occurrences in the English source and those that corresponded to a match in the Portuguese translation. Texts in the Brazilian variant are identified in bold in t
he table.
\par Table 6 shows that some pairs are more useful than others, and also how this may vary for individual texts. Although we have to leave a detailed discussion of these results for another forum, we provide here some preliminary comments.
\par One case that is worth discussing is }{\i I}{/}{\i eu}{, which is interesting in that }{\i eu}{ is considerably more rare in Portuguese (being a null subject language) than }{\i I}{
 in English. One can, however, notice that there is a considerable difference in utility (or translation correspon
dence) from text to text, and even note that, as expected, in the Brazilian variant there is a less marked tendency to dispose of the personal pronoun than in European Portuguese. 
\par The opposite happens with dashes (coded }{\i &mdash;}{) in terms of relative frequ
encies: dashes are used much more frequently in Portuguese than in English. It is interesting to argue for a different behaviour of the aligner in such cases, i.e., a different weighting of pairs related to criteria such as these. (The lower the frequency
 of the target item, the higher the probability that, being there, it is a translation of the source term.) 
\par }\pard\plain \s19\qc\sb120\sa120\widctlpar\adjustright \fs20\cgrid {\i Table 6.}{ The distribution of some frequent anchor pairs
\par }\trowd \trqc\trleft336\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx1207\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1676\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx2265\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2734\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx3203\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3672\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx4261\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4730\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx5199\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5668\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx6137\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6606\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx7075\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7544\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx8013\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8482\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx8951\pard\plain \s23\widctlpar\intbl\adjustright \lang2057\cgrid {pair\cell }\pard\plain \qc\widctlpar\intbl\adjustright \lang2057\cgrid {\b AB\cell AH\cell AT\cell }{BC\cell D1\cell D2\cell FF\cell }\pard\plain 
\s3\qc\keepn\widctlpar\intbl\outlinelevel2\adjustright \b\lang2057\cgrid {JB\cell }\pard\plain \qc\widctlpar\intbl\adjustright \lang2057\cgrid {JB'\cell JH\cell MA\cell }\pard\plain \s3\qc\keepn\widctlpar\intbl\outlinelevel2\adjustright \b\lang2057\cgrid 
{NG\cell }\pard\plain \qc\widctlpar\intbl\adjustright \lang2057\cgrid {PD\cell RD\cell ST\cell }\pard\plain \s3\qc\keepn\widctlpar\intbl\outlinelevel2\adjustright \b\lang2057\cgrid {WB\cell }\pard\plain \widctlpar\intbl\adjustright \lang2057\cgrid {\row 
}\trowd \trqc\trleft336\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx1207\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1676\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx2265\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2734\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx3203\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3672\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx4261\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4730\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx5199\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5668\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx6137\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6606\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx7075\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7544\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx8013\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8482\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr
\brdrs\brdrw10 \cltxlrtb \cellx8951\pard \widctlpar\intbl\adjustright {&mdash
\par &mdash
\par \cell }\pard \qc\widctlpar\intbl\adjustright {57
\par 247
\par 16\cell 65
\par 454
\par 25\cell 53
\par 343
\par 46\cell 66
\par 305
\par 50\cell 66
\par 236
\par 22\cell 36
\par 507
\par 22\cell 8
\par 12
\par 10\cell 37
\par 47
\par 30\cell 37
\par 130
\par 33\cell 8
\par 36
\par 7\cell 7
\par 33
\par 6\cell 130
\par 127
\par 107\cell 17
\par 210
\par 7\cell 436
\par 445
\par 426\cell 20
\par 389
\par 4\cell 36
\par 234
\par 32\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {'s
\par \'e9, est\'e1
\par \cell }\pard \qc\widctlpar\intbl\adjustright {82
\par 504
\par 30\cell 75
\par 520
\par 35\cell 38
\par 300
\par 12\cell 2
\par 304
\par 2\cell 40
\par 320
\par 21\cell 71
\par 432
\par 27\cell 11
\par 100
\par 3\cell 98
\par 576
\par 55\cell 92
\par 608
\par 57\cell 0
\par 236
\par 0\cell 69
\par 584
\par 33\cell 50
\par 264
\par 29\cell 67
\par 504
\par 34\cell 62
\par 316
\par 33\cell 39
\par 504
\par 15\cell 16
\par 144
\par 11\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {I
\par eu
\par \cell }\pard \qc\widctlpar\intbl\adjustright {238
\par 105
\par 87\cell 171
\par 61
\par 40\cell 73
\par 32
\par 19\cell 84
\par 20
\par 17\cell 43
\par 11
\par 8\cell 115
\par 39
\par 19\cell 39
\par 1
\par 1\cell 316
\par 85
\par 62\cell 316
\par 122
\par 104\cell 11
\par 4
\par 3\cell 217
\par 37
\par 32\cell 110
\par 53
\par 43\cell 124
\par 35
\par 22\cell 322
\par 109
\par 94\cell 80
\par 43
\par 30\cell 205
\par 71
\par 63\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {be
\par ser, estar
\par \cell }\pard \qc\widctlpar\intbl\adjustright {65
\par 172
\par 18\cell 92
\par 116
\par 17\cell 41
\par 132
\par 14\cell 21
\par 60
\par 5\cell 65
\par 152
\par 13\cell 66
\par 120
\par 9\cell 64
\par 84
\par 9\cell 54
\par 172
\par 19\cell 54
\par 248
\par 29\cell 48
\par 160
\par 15\cell 38
\par 116
\par 11\cell 61
\par 172
\par 24\cell 80
\par 172
\par 20\cell 24
\par 68
\par 3\cell 61
\par 220
\par 8\cell 19
\par 60
\par 11\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {been
\par sido, estado\cell }\pard \qc\widctlpar\intbl\adjustright {62
\par 64
\par 13\cell 45
\par 36
\par 0\cell 33
\par 40
\par 7\cell 21
\par 64
\par 7\cell 53
\par 72
\par 10\cell 42
\par 48
\par 4\cell 62
\par 28
\par 5\cell 28
\par 40
\par 4\cell 28
\par 64
\par 11\cell 35
\par 72
\par 10\cell 13
\par 32
\par 4\cell 41
\par 36
\par 5\cell 55
\par 100
\par 18\cell 19
\par 68
\par 6\cell 45
\par 84
\par 7\cell 24
\par 12
\par 2\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {could
\par }{\fs16 podia.*, etc. \cell }\pard \qc\widctlpar\intbl\adjustright {29
\par 132
\par 12\cell 19
\par 120
\par 10\cell 48
\par 216
\par 28\cell 18
\par 96
\par 8\cell 33
\par 114
\par 7\cell 41
\par 66
\par 6\cell 45
\par 168
\par 18\cell 25
\par 126
\par 10\cell 25
\par 138
\par 9\cell 54
\par 180
\par 24\cell 13
\par 42
\par 6\cell 33
\par 222
\par 20\cell 68
\par 180
\par 25\cell 46
\par 150
\par 19\cell 32
\par 108
\par 9\cell 21
\par 90
\par 10\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {is
\par \'e9, est\'e1
\par \cell }\pard \qc\widctlpar\intbl\adjustright {65
\par 504
\par 34\cell 51
\par 520
\par 29\cell 14
\par 300
\par 12\cell 41
\par 304
\par 26\cell 29
\par 320
\par 13\cell 47
\par 432
\par 17\cell 25
\par 100
\par 7\cell 60
\par 576
\par 38\cell 60
\par 608
\par 41\cell 53
\par 236
\par 28\cell 130
\par 584
\par 69\cell 25
\par 264
\par 17\cell 45
\par 504
\par 24\cell 19
\par 316
\par 15\cell 35
\par 504
\par 16\cell 21
\par 144
\par 11\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\intbl\adjustright {his
\par seu*, dele*\cell }\pard \qc\widctlpar\intbl\adjustright {81
\par 220
\par 19\cell 81
\par 272
\par 12\cell 167
\par 488
\par 64\cell 142
\par 364
\par 52\cell 66
\par 264
\par 24\cell 57
\par 212
\par 13\cell 157
\par 268
\par 37\cell 59
\par 124
\par 7\cell 59
\par 364
\par 23\cell 158
\par 360
\par 61\cell 37
\par 136
\par 12\cell 95
\par 488
\par 44\cell 153
\par 496
\par 60\cell 95
\par 272
\par 25\cell 99
\par 208
\par 15\cell 94
\par 224
\par 26\cell }\pard \widctlpar\intbl\adjustright {\row }\trowd \trqc\trleft336\trbrdrt\brdrs\brdrw10 \trbrdrl\brdrs\brdrw10 \trbrdrb\brdrs\brdrw10 \trbrdrr\brdrs\brdrw10 \trbrdrh\brdrs\brdrw10 \trbrdrv\brdrs\brdrw10 \clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1207\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx1676\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2265\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx2734\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3203\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx3672\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4261\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx4730\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5199\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx5668\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6137\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx6606\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7075\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx7544\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8013\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8482\clvertalt\clbrdrt\brdrs\brdrw10 \clbrdrl
\brdrs\brdrw10 \clbrdrb\brdrs\brdrw10 \clbrdrr\brdrs\brdrw10 \cltxlrtb \cellx8951\pard \widctlpar\intbl\adjustright {there
\par l\'e1, ali
\par \cell }\pard \qc\widctlpar\intbl\adjustright {52
\par 144
\par 14\cell 54
\par 88
\par 11\cell 42
\par 84
\par 5\cell 23
\par 32
\par 2\cell 47
\par 124
\par 14\cell 44
\par 120
\par 7\cell 43
\par 28
\par 2\cell 38
\par 172
\par 11\cell 38
\par 68
\par 10\cell 47
\par 36
\par 8\cell 64
\par 80
\par 8\cell 58
\par 112
\par 21\cell 51
\par 140
\par 10\cell 79
\par 180
\par 18\cell 23
\par 152
\par 3\cell 31
\par 68
\par 6\cell }\pard \widctlpar\intbl\adjustright {\row }\pard \widctlpar\adjustright {
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {It is, however, also possible to see great variation without having either a significant frequency reduction or increase of the corresponding terms in the two languages, as the cases of }{\i been}{
 and }{\i sido/estado}{ show. More detailed studies of the subtleties of the translation into Portuguese of }{\i could}{ and of }{\i there}{ can be found, respectively, in Santos (1998a) and Ebeling & Oksefjell (1998).
\par }\pard\plain \s2\sb240\sa60\keepn\widctlpar\outlinelevel1\adjustright \b\f4\fs28\lang2057\cgrid {5.3 Looking at some instances of successful matches
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {One thing worth noticing is that, even though much of the data presented above could be explained with some linguistic ingenuity, it is not t
he case that all instances that were counted as successful matches are actually linguistically motivated. In fact, it was easy to detect three interesting situations from a crude inspection of the results:
\par }\pard\plain \s37\qj\fi709\sa120\widctlpar\adjustright \cgrid {First, even though the result might be correct, some of the resulting pairs "succeeded" without any sort of linguistic motivation; cf. (1) - (4):
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (1)\tab}}\pard\plain \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright 
\fs22\cgrid {little / pequen.*?: }{\i \'c0 sua frente, pelo lado esquerdo do caminho, ia um grupo de crian\'e7as, a mais velha a empurrar um carrinho, com duas crian\'e7as mais pequenas, uma de cada lado, agarradas ao var\'e3o.}{
 (Translation of: Ahead of him, trudging along on the left of the path, was a little group of children, the eldest girl wheeling a pushchair with two smaller children, one each side of her, clutching the bars, PDJ3)
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (2)\tab}}\pard \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright {be 
/ ((ser)|(estar)): }{\i L\'e1 um }{\b\i ser}{\i  }{('being', n.)}{\i  humano sempre ser\'e1 um estranho, jamais "pertencer\'e1".}{ (Translation of: In there, one will always }{\b be}{ a stranger, will never "belong", ABR1) 
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (3)\tab}}\pard \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright {
look.*? / olh.*?: }{\i N\'e3o h\'e1 quebra-luz, apenas a l\'e2mpada por cima que d\'e1 \'e0 minha cara um aspecto p\'e1lido e doentio, com }{\b\i olheiras}{\i .}{
(Translation of: There 's no shade on the light, just a bare bulb overhead, which makes my face }{\b look}{ pallid and ill, with circles under the eyes, MA1.)
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (4)\tab}}\pard \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright {
couple.*? / par.*?: }{\i Depois de alguns anos }{\b\i para}{\i  }{('for')}{\i  se estabelecer, ele come\'e7a a escrever. }{(Translation of: After a }{\b couple}{ of years to settle down he begins to write again, ABR1).
\par }\pard\plain \s37\qj\fi709\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {Secondly, it was also noted that the use of the Kleene star, instead of a more rigid morphologically closed list, allowed for 
correct matches which would certainly not have been listed by a linguist (due to considerations of frequency), as is the case of (5)-(6). Incidentally, a particularly valuable feature of the anchor list format was the ease of specifying cross-categorial t
ranslations (i.e., part-of-speech change in the translation).
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (5)\tab}}\pard\plain \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright 
\fs22\cgrid {buil.*? / constru.*?: }{\i Terminada a obra, os }{\b\i construtores}{\i  ser\'e3o mortos. }{(Translation of: Once the job is finished the }{\b builders}{ are killed, ABR1.)
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (6)\tab}}\pard \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright {
crim.*? / crim.*?: }{\i Ent\'e3o, perto do final de 1347, uma pequena frota de cerca de uma d\'fazia de galeras genovesas chegou \'e0 Sic\'edlia vinda de um lugar distante, talvez da }{\b\i Crim\'e9ia}{\i , e em poucos dias a popula\'e7\'e3
o de Messina come\'e7ou a morrer \'e0s centenas.}{ (Translation of: Then, towards the end of 1347, a small fleet of about a dozen Genoan galleys arrived in Sicily from somewhere far away, perhaps the }{\b Crimea}{
, and within a few days the people of Messina began to die in their hundreds, ABR1)
\par }\pard\plain \s37\qj\fi709\sa120\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {Thirdly, example (7) displays a match which works despite the linguistic data i
t is supposed to mirror and/or despite the inaccurate translation. In fact, }{\i earlier}{ in (7) was translated simply by }{\i cedo}{ ('early'), leaving out the comparison (a literal translation would use }{\i mais cedo}{
), but the fact that one has chosen to simply list the absolute form made the anchor pair more useful.
\par {\pntext\pard\plain\s35 \fs22\cgrid \hich\af0\dbch\af0\loch\f0 (7)\tab}}\pard\plain \s35\qj\fi-720\li720\sa60\widctlpar\jclisttab\tx720{\*\pn \pnlvlbody\ilvl0\ls24\pnrnot0\pndec\pnstart1\pnindent720\pnhang{\pntxtb (}{\pntxta )}}\ls24\adjustright 
\fs22\cgrid {earli.*? / cedo: }{\i Ela trabalhava mais que qualquer um, acordava }{\b\i cedo }{('early')}{\i , era a que ia dormir mais tarde.}{ (Translation of: She worked harder than anybody else, got up }{\b earlier}{, came to bed long after the 
others, ABR1)
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {
In conclusion, it was possible to detect several pairs that led to "successes" without having been thought of as pairs beforehand. In some cases this increases performance, in others it diminishes it. Without looking at every pair that counte
d as a success, it is difficult to estimate the actual percentage of these cases.
\par The overall performance of the aligner did not appear significantly diminished, though, since, as mentioned above, there can be many possible unrelated "successes" for each s
-unit. This rather illustrates that the TCA's approach, while not 100% linguistically motivated, seems to compensate on accounts of both simplicity and robustness.
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 6.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid { Concluding remarks
\par }\pard\plain \s34\qj\widctlpar\adjustright \cgrid {Even though it was already known from the beginning that the TCA had been of extreme help in aligning the English-Portuguese texts of the ENPC, we think that the evaluation performed was interesting.}{
\lang2070 
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {On the one hand, it measures the actual consequences of using this particular program for a given language pair, which is someth
ing that had not been done before. It is then up to the individual user to decide whether s/he should use this tool in his or her project.
\par On the other hand, this investigation also gives origin to some reflections concerning the evaluation process proper and the role of language dependency in NLP tools:
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li720\widctlpar\jclisttab\tx720{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
One of the undeniable conclusions of the present study is, in fact, that it is a time-consuming and complex task to evaluate a particular program, and that it is often necessary to create auxiliary tools to
 evaluate it. It is, therefore, understandable that, in most cases, tool developers have no time to implement a thorough evaluation of the tools they develop because they employ their time in improving them (adding more functionalities, for example), not 
e
valuating them. It is, therefore, understandable that, in most cases, tool developers have no time to implement a thorough evaluation of the tools they develop.  Furthermore, it is often unclear whether the time used in this kind of evaluation is worthwhi
le, given that the program has a satisfactory behaviour for the task it was meant to perform.
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li720\widctlpar\jclisttab\tx720{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
While adapting a particular language-dependent tool for other typologically similar languages, it seems that performance degradation is rarely so severe that it pr
ohibits reuse of the tool. It is, however, seldom quantified what the changes are, and how well the "new" tool works when adapted to the new language (in this case, new pair of languages). That is why the present study may be interesting beyond the partic
ular tool and pair of languages considered.
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li720\widctlpar\jclisttab\tx720{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
Our preliminary conclusion is that even language-dependent tools }{\cf11 rely}{ to a lesser extent on languages than one would expect. Therefore, it is often feasible (and useful from an engineering}{\cf11  }{point of view) to adapt 
that program to similar languages without considerable loss of performance. But see Santos (in press) for a more detailed discussion of language dependency and its methodological consequences.
\par }\pard\plain \s16\qj\fi709\widctlpar{\*\pn \pnlvlcont\ilvl0\ls0\pnrnot0\pndec }\adjustright \cgrid {As to the particular work reported here, we note that there are
 two other ways in which we could proceed for a thorough evaluation of the TCA for the English-Portuguese language pair: 
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard\plain \s36\qj\fi-360\li720\widctlpar\jclisttab\tx720{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright \cgrid {
The first way would be to experiment with the constitution of the anchor list, creating new anchor lists, including for instance only 
the 20 most frequent pairs, or only the translations of the most frequent English content words, and compare the performances obtained.
\par {\pntext\pard\plain\s36 \f3\cgrid \loch\af3\dbch\af0\hich\f3 \'b7\tab}}\pard \s36\qj\fi-360\li720\widctlpar\jclisttab\tx720{\*\pn \pnlvlblt\ilvl0\ls27\pnrnot0\pnf3\pnstart1\pnindent360\pnhang{\pntxtb \'b7}}\ls27\adjustright {
The second way would be to apply new texts and their translations to the TCA, and see whether this would result in any appreciable decrease in performance.
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {Unfortunately, both tasks will have to wait for another occasion, since they amount to considerable work.}{\cs26\super \chftn {\footnote\ftnalt \pard\plain \s25\qj\widctlpar\adjustright 
\fs20\lang2057\cgrid {\cs26\super \chftn }{ In order t
o compare the performance of the system with a revised version, one has to manually proofread the results first. For the case of the ENPC we had access to the proofread files, which is obviously not the general case.}}}{
 Another revealing experiment would be to mimick the whole evaluation process for the English-Norwegian pair and compare the results.
\par {\pntext\pard\plain\s1 \b\fs28\cgrid \hich\af0\dbch\af0\loch\f0 7.\tab}}\pard\plain \s1\fi-357\li357\sb240\sa120\keepn\widctlpar\jclisttab\tx360{\*\pn \pnlvlbody\ilvl0\ls21\pnrnot0\pndec\pnstart1\pnindent360\pnhang{\pntxta .}}
\ls21\outlinelevel0\adjustright \b\fs28\cgrid {Acknowledgements
\par }\pard\plain \s16\qj\fi709\widctlpar\adjustright \cgrid {
We are grateful to Knut Hofland for making his program widely available and for answering our questions about it. In addition, the present paper was considerably improved by Stig Johansson's comments on both structure a
nd content. We also thank Ana Frankenberg-Garcia for commenting on a previous draft.}{\cf6 
\par }{\sect }\sectd \sbknone\linex0\headery709\footery709\colsx709\endnhere\sectdefaultcl \pard\plain \s1\sb240\sa120\keepn\widctlpar\outlinelevel0\adjustright \b\fs28\cgrid {References
\par }\pard\plain \s28\qj\fi-567\li567\ri227\sa120\widctlpar\tx993\adjustright \f4\lang1044\cgrid {Ebeling, J. & Oksefjell, S. 1998. On the translation of English there-sentences into Norwegian and Portuguese. What does a translation corpus tell us?. In Ydst
ie, J. T. & Wolleb\'e6k, A.C. (eds.), }{\i Working Papers in Applied Linguistics 4/98}{, Oslo: Department of linguistics, Faculty of Arts, University of Oslo. 188-206.
\par Hofland, K. 1996. A program for aligning English and Norwegian sentences. In Hockey, S, Ide, N. & Perissinotto, G. (eds.), }{\i Research in Humanities Computing}{. Oxford: Oxford University Press. 165-178.
\par Hofland, K. & Johansson, S. 1998. The Translation Corpus Aligner: A program for automatic alignment of parallel texts. In Johansson, S. & Oksefjell, S. (eds.), }{\i Corpora and Cross-linguistic Research: Theory, Method, and Case Studies}{
. Amsterdam: Rodopi. 87-100.
\par Johansson, S, Ebeling, J. & Hofland, K. 1996. Coding and aligning the English-Norwegian Parallel Corpus. In Aijmer, K., Altenberg, B. & Johansson, M. (eds.). }{\i 
Languages in contrast. Papers from a symposium on text-based cross-linguistic studies in Lund}{ (Lund, 4-5 March 1994). Lund: Lund University Press. 87-112.
\par Johansson, S., Ebeling, J. & Oksefjell, S. 1999. English-Norwegian Parallel Corpus: Manual. Oslo: Department of British and American Studies, University of Oslo, }{\field{\*\fldinst { HYPERLINK http://www.hf.uio.no/iba/prosjekt/ENPCmanual.html }{
{\*\datafield 
00d0c9ea79f9bace118c8200aa004ba90b02000000170000003200000068007400740070003a002f002f007700770077002e00680066002e00750069006f002e006e006f002f006900620061002f00700072006f0073006a0065006b0074002f0045004e00500043006d0061006e00750061006c002e00680074006d006c00
0000e0c9ea79f9bace118c8200aa004ba90b6400000068007400740070003a002f002f007700770077002e00680066002e00750069006f002e006e006f002f006900620061002f00700072006f0073006a0065006b0074002f0045004e00500043006d0061006e00750061006c002e00680074006d006c000000}}
}{\fldrslt {\cs22\ul\cf2 http://www.hf.uio.no/iba/prosjekt/ENPCmanual.html}}}{.
\par Santos, D. 1998a. Perception verbs in English and Portuguese. In Johansson, S. & Oksefjell, S. (eds.),}{\i  Corpora and Cross-linguistic Research: Theory, Method, and Case Studies}{. Amsterdam: Rodopi. 319-342.
\par Santos, D. 1998b. Punctuation and multilinguality: Reflections from a language engineering perspective. In J. T. Ydstie & A.C. Wolleb\'e6k (eds.), }{\i Working Papers in Applied Linguistics 4/98}{
, Oslo: Department of linguistics, Faculty of Arts, University of Oslo.138-160.
\par Santos, D. In press. Towards language-dependent applications,}{\i  Machine Translation}{ 14 (1999).
\par Sperberg-McQueen, M. C. M. & Burnard, L. (eds). 1994. }{\i Guidelines for electronic text encoding and interchange. TEI P3}{
. Chicago & Oxford: Association for Computers and the Humanities / Association for Computational Linguistics / Association for Literary and Linguistic Computing.
\par }}