Fichier de travail (INPUT) : ./DUMP-TEXT/2-3.txt

Encodage utilisé (INPUT) : utf-8

Forme recherchée : translation|traduction|机器

Ligne n°45 : ... posts rolling in on social media. Apparently Google Translate, the
• Ligne n°46 : company’s popular machine-translation service, had suddenly and almost
Ligne n°47 : immeasurably improved. Rekimoto visited Translate himself and began to ...

Ligne n°52 : ... compared a few sentences from two published versions of “The Great
• Ligne n°53 : Gatsby,” Takashi Nozaki’s 1957 translation and Haruki Murakami’s more
Ligne n°54 : recent iteration, with what this new Google Translate was able to ...

Ligne n°54 : ... recent iteration, with what this new Google Translate was able to
• Ligne n°55 : produce. Murakami’s translation is written “in very polished Japanese,”
Ligne n°56 : Rekimoto explained to me later via email, but the prose is ...

Ligne n°56 : ... Rekimoto explained to me later via email, but the prose is
• Ligne n°57 : distinctively “Murakami-style.” By contrast, Google’s translation —
Ligne n°58 : despite some “small unnaturalness” — reads to him as “more ...

Ligne n°98 : ... on Twitter, and over the next few hours thousands of people broadcast
• Ligne n°99 : their own experiments with the machine-translation service. Some were
Ligne n°100 : successful, others meant mostly for comic effect. As dawn broke over ...

Ligne n°160 : ... behind him appeared a graph whose steep curve indicated a recent
• Ligne n°161 : fivefold increase in translations between Arabic and German. (It was
Ligne n°162 : also close to Pichai’s own heart. He grew up in India, a land divided ...

Ligne n°169 : ... States but in Europe and Asia as well: The rollout included
• Ligne n°170 : translations between English and Spanish, French, Portuguese, German,
Ligne n°171 : Chinese, Japanese, Korean and Turkish. The rest of Translate’s ...

Ligne n°327 : ... View over nine months, and it explains the transformation of machine
• Ligne n°328 : translation. The second story, the story of Google Brain and its many
Ligne n°329 : competitors, takes place in Silicon Valley over five years, and it ...

Ligne n°1045 : ... into how machines might generate art and music. It also cleared the way
• Ligne n°1046 : toward an instrumental task like machine translation. Hinton told me he
Ligne n°1047 : thought at the time that this follow-up work would take at least five ...


• Ligne n°1052 : Le’s paper showed that neural translation was plausible, but he had
Ligne n°1053 : used only a relatively small public data set. (Small for Google, that ...

Ligne n°1121 : ... 10 years — with a neural network. The old system worked the way all
• Ligne n°1122 : machine translation has worked for about 30 years: It sequestered each
Ligne n°1123 : successive sentence fragment, looked up those words in a large ...

Ligne n°1126 : ... make sense. The approach is called “phrase-based statistical machine
• Ligne n°1127 : translation,” because by the time the system gets to the next phrase,
Ligne n°1128 : it doesn’t know what the last one was. This is why Translate’s output ...

Ligne n°1137 : ... more than an expected incremental bump. But there was a case to be made
• Ligne n°1138 : that human-quality machine translation is not only a short-term
Ligne n°1139 : necessity but also a development very likely, in the long term, to ...

Ligne n°1144 : ... majority of market share in search-engine traffic belonged to its
• Ligne n°1145 : competitor Baidu — or India, decent machine translation would be an
Ligne n°1146 : indispensable part of the infrastructure. Baidu itself had published a ...

Ligne n°1146 : ... indispensable part of the infrastructure. Baidu itself had published a
• Ligne n°1147 : pathbreaking paper about the possibility of neural machine translation
Ligne n°1148 : in July 2015. ...


• Ligne n°1150 : And in the more distant, speculative future, machine translation was
Ligne n°1151 : perhaps the first step toward a general computational facility with ...

Ligne n°1165 : ... gear. He knew that various people in various places at Google and
• Ligne n°1166 : elsewhere had been trying to make neural translation work — not in a
Ligne n°1167 : lab but at production scale — for years, to little avail. ...

Ligne n°1193 : ... It was a challenge Schuster couldn’t resist. The benchmark metric to
• Ligne n°1194 : evaluate machine translation is called a BLEU score, which compares a
Ligne n°1195 : machine translation with an average of many reliable human ...

Ligne n°1194 : ... evaluate machine translation is called a BLEU score, which compares a
• Ligne n°1195 : machine translation with an average of many reliable human
Ligne n°1196 : translations. At the time, the best BLEU scores for English-French were ...

Ligne n°1195 : ... machine translation with an average of many reliable human
• Ligne n°1196 : translations. At the time, the best BLEU scores for English-French were
Ligne n°1197 : in the high 20s. An improvement of one point was considered very good; ...


• Ligne n°1218 : Until then, the neural-translation team had been only three people —
Ligne n°1219 : Schuster, Wu and Chen — but with Hughes’s support, the broader team ...

Ligne n°1255 : ... Hemingway. Some of this is in the public domain: The original Rosetta
• Ligne n°1256 : Stone of statistical machine translation was millions of pages of the
Ligne n°1257 : complete bilingual records of the Canadian Parliament. Much of it, ...

Ligne n°1258 : ... however, was culled from 10 years of collected data, including human
• Ligne n°1259 : translations that were crowdsourced from enthusiastic respondents. The
Ligne n°1260 : team had in their storehouse about 97 million unique English “words.” ...

Ligne n°1265 : ... Google’s Quoc Le (right), whose work demonstrated the plausibility of
• Ligne n°1266 : neural translation, with Mike Schuster, who helped apply that work to
Ligne n°1267 : Google Translate.CreditBrian Finke for The New York Times ...

Ligne n°1275 : ... direction. The network was very sensitive to the data it was trained
• Ligne n°1276 : on. As Hughes put it to me at one point: “The neural-translation system
Ligne n°1277 : is learning everything it can. It’s like a toddler. ‘Oh, Daddy says ...

Ligne n°1281 : ... thing was fast and reliable enough that their users wouldn’t notice. In
• Ligne n°1282 : February, the translation of a 10-word sentence took 10 seconds. They
Ligne n°1283 : could never introduce anything that slow. The Translate team began to ...

Ligne n°1284 : ... conduct latency experiments on a small percentage of users, in the form
• Ligne n°1285 : of faked delays, to identify tolerance. They found that a translation
Ligne n°1286 : that took twice as long, or even five times as long, wouldn’t be ...

Ligne n°1353 : ... They weren’t just debugging the model; they were debugging the chip.
• Ligne n°1354 : The neural-translation project would be proof of concept for the whole
Ligne n°1355 : infrastructural investment. ...


• Ligne n°1416 : Google’s neural translation was at last working. By the time of the
Ligne n°1417 : party, the company’s Chinese-English test had already processed 18 ...

Ligne n°1424 : ... When word began to spread, over the following weeks, that Google had
• Ligne n°1425 : introduced neural translation for Chinese to English, some people
Ligne n°1426 : speculated that it was because that was the only language pair for ...

Ligne n°1441 : ... important to him to underline the fact, he began, that the neural
• Ligne n°1442 : translation project itself represented a “collaboration between groups
Ligne n°1443 : that spoke different languages.” ...


• Ligne n°1445 : Their neural-translation project, he continued, represented a “step
Ligne n°1446 : function forward” — that is, a discontinuous advance, a vertical leap ...

Ligne n°1446 : ... function forward” — that is, a discontinuous advance, a vertical leap
• Ligne n°1447 : rather than a smooth curve. The relevant translation had been not just
Ligne n°1448 : between the two teams but from theory into reality. He raised a plastic ...

Ligne n°1476 : ... or the claims made on its behalf, implicates the question of
• Ligne n°1477 : translation. The Chinese Room argument was proposed in 1980 by the
Ligne n°1478 : Berkeley philosopher John Searle. In Searle’s thought experiment, a ...

Ligne n°1530 : ... mere statistical prediction, a glorified weather forecast. Even if
• Ligne n°1531 : neural translation attained perfect functionality, it would reveal
Ligne n°1532 : nothing profound about the underlying nature of language. It could ...