Fichier de travail (INPUT) : ./DUMP-TEXT/2-41.txt
Encodage utilisé (INPUT) : utf-8
Forme recherchée : translation|traduction|机器
_________________________________________________________________________________________________
- Ligne n°61 : Machine translation works well for sentences but turns out to falter at the
Ligne n°62 : document level, computational linguists have found. ...
Ligne n°82 : ... ticker tape, but in recent months computational linguists have begun to- Ligne n°83 : claim that neural machine translation now matches the performance of
Ligne n°84 : human translators. ...
Ligne n°109 : ... machines lag significantly behind humans, they say.- Ligne n°110 : [machine-translation.png?sw=600&cx=0&cy=0&cw=1420&ch=53
Ligne n°111 : 5] ...- Ligne n°113 : At issue is how machine translation should be evaluated. This is
Ligne n°114 : currently done on two measures: adequacy and fluency. The adequacy of a ...
Ligne n°114 : ... currently done on two measures: adequacy and fluency. The adequacy of a- Ligne n°115 : translation is determined by professional human translators who read
Ligne n°116 : both the original text and the translation to see how well it expresses ...
Ligne n°115 : ... translation is determined by professional human translators who read- Ligne n°116 : both the original text and the translation to see how well it expresses
Ligne n°117 : the meaning of the source. Fluency is judged by monolingual readers who ...
Ligne n°117 : ... the meaning of the source. Fluency is judged by monolingual readers who- Ligne n°118 : see only the translation and determine how well it is expressed in
Ligne n°119 : English. ...
Ligne n°122 : ... But according to Laubli and co, the current protocol only compares- Ligne n°123 : translations at the sentence level, whereas humans also evaluate text
Ligne n°124 : at the document level. ...
Ligne n°129 : ... translated over 100 news articles written in Chinese into English. The- Ligne n°130 : examiners rated each translation for adequacy and fluency at the
Ligne n°131 : sentence level but, crucially also at the level of the entire document. ...
Ligne n°139 : ... However, when it comes to evaluating the entire document, human- Ligne n°140 : translations are rated as more adequate and more fluent than machine
Ligne n°141 : translations. “Human raters assessing adequacy and fluency show a ...
Ligne n°140 : ... translations are rated as more adequate and more fluent than machine- Ligne n°141 : translations. “Human raters assessing adequacy and fluency show a
Ligne n°142 : stronger preference for human over machine translation when evaluating ...
Ligne n°141 : ... translations. “Human raters assessing adequacy and fluency show a- Ligne n°142 : stronger preference for human over machine translation when evaluating
Ligne n°143 : documents as compared to isolated sentences,” they say. ...
Ligne n°145 : ... The researchers think they know why. “We hypothesise that- Ligne n°146 : document-level evaluation unveils errors such as mistranslation of an
Ligne n°147 : ambiguous word, or errors related to textual cohesion and coherence, ...- Ligne n°158 : This suggests that the way machine translation is evaluated needs to
Ligne n°159 : evolve away from a system where machines consider each sentence in ...- Ligne n°162 : “As machine translation quality improves, translations will become
- Ligne n°162 : “As machine translation quality improves, translations will become
Ligne n°163 : harder to discriminate in terms of quality, and it may be time to shift ...
Ligne n°164 : ... towards document-level evaluation, which gives raters more context to- Ligne n°165 : understand the original text and its translation, and also exposes
Ligne n°166 : translation errors related to discourse phenomena which remain ...
Ligne n°165 : ... understand the original text and its translation, and also exposes- Ligne n°166 : translation errors related to discourse phenomena which remain
Ligne n°167 : invisible in a sentence-level evaluation,” say Laubli and co. ...- Ligne n°169 : That change should help machine translation improve. Which means it is
Ligne n°170 : still set to surpass human translation—just not yet. ...
Ligne n°169 : ... That change should help machine translation improve. Which means it is- Ligne n°170 : still set to surpass human translation—just not yet.