Fichier de travail (INPUT) : ./DUMP-TEXT/2-4.txt
Encodage utilisé (INPUT) : utf-8
Forme recherchée : translation|traduction|机器
_________________________________________________________________________________________________
Ligne n°9 : ... * 2Speech recognition: I hear you- Ligne n°10 : * 3Machine translation: Beyond Babel
Ligne n°11 : * 4Meaning and machine intelligence: What are you talking about? ...- Ligne n°31 : Computers have got much better at translation, voice recognition and
Ligne n°32 : speech synthesis, says Lane Greene. But they still don’t understand the ...- Ligne n°64 : Speech recognition has made remarkable advances. Machine translation,
Ligne n°65 : too, has gone from terrible to usable for getting the gist of a text, ...
Ligne n°85 : ... Sciences. In the period leading up to this, scholars had been promising- Ligne n°86 : automatic translation between languages within a few years.
Ligne n°88 : ... But the report was scathing. Reviewing almost a decade of work on- Ligne n°89 : machine translation and automatic speech recognition, it concluded that
Ligne n°90 : the time had come to spend money “hard-headedly toward important, ...
Ligne n°124 : ... Many early approaches to language technology—and particularly- Ligne n°125 : translation—got stuck in a conceptual cul-de-sac: the rules-based
Ligne n°126 : approach. In translation, this meant trying to write rules to analyse ...
Ligne n°125 : ... translation—got stuck in a conceptual cul-de-sac: the rules-based- Ligne n°126 : approach. In translation, this meant trying to write rules to analyse
Ligne n°127 : the text of a sentence in the language of origin, breaking it down into ...
Ligne n°139 : ... has learned to make its best guess about a previously unseen text. In- Ligne n°140 : machine translation, the software scans millions of words already
Ligne n°141 : translated by humans, again looking for patterns. In speech ...
Ligne n°148 : ... poised to benefit. People who had been put off by the hilariously- Ligne n°149 : inappropriate translations offered by online tools like BabelFish began
Ligne n°150 : to have more faith in Google Translate. Apple persuaded millions of ...- Ligne n°370 : Machine translation: Beyond Babel
- Ligne n°372 : Computer translations have got strikingly better, but still need human
Ligne n°373 : input ...
Ligne n°378 : ... civilisations generally requires some kind of device to allow them to- Ligne n°379 : talk. High-quality automated translation seems even more magical than
Ligne n°380 : other kinds of language technology because many humans struggle to ...- Ligne n°383 : The idea has been around since the 1950s, and computerised translation
Ligne n°384 : is still known by the quaint moniker “machine translation” (MT). It ...
Ligne n°383 : ... The idea has been around since the 1950s, and computerised translation- Ligne n°384 : is still known by the quaint moniker “machine translation” (MT). It
Ligne n°385 : goes back to the early days of the cold war, when American scientists ...
Ligne n°396 : ... a demonstration in New York on January 7th 1954 and proudly produced 60- Ligne n°397 : automated translations, including that of “Mi pyeryedayem mislyi
Ligne n°398 : posryedstvom ryechyi,” which came out correctly as “We transmit ...
Ligne n°409 : ... by private companies. The most notable of them was Systran, which- Ligne n°410 : provided rough translations, mostly to America’s armed forces.
Ligne n°426 : ... “the pen is in the box” and “the box is in the pen” would require- Ligne n°427 : different translations for “pen”: any pen big enough to hold a box
Ligne n°428 : would have to be an animal enclosure, not a writing instrument. ...
Ligne n°435 : ... use statistical probabilities rather than rules devised by humans for- Ligne n°436 : translation. Statistical, “phrase-based” machine translation, like
- Ligne n°436 : translation. Statistical, “phrase-based” machine translation, like
Ligne n°437 : speech recognition, needed training data to learn from. Candide used ...
Ligne n°439 : ... in French and English, providing a huge amount of data for that time.- Ligne n°440 : The phrase-based approach would ensure that the translation of a word
Ligne n°441 : would take the surrounding words properly into account. ...
Ligne n°446 : ... goal of indexing the entire internet, decided to use those data to- Ligne n°447 : train its translation engines; in 2007 it switched from a rules-based
Ligne n°448 : engine (provided by Systran) to its own statistics-based system. To ...
Ligne n°449 : ... build it, Google trawled about a trillion web pages, looking for any- Ligne n°450 : text that seemed to be a translation of another—for example, pages
Ligne n°451 : designed identically but with different words, and perhaps a hint such ...
Ligne n°457 : ... Training on parallel texts (which linguists call corpora, the plural of- Ligne n°458 : corpus) creates a “translation model” that generates not one but a
Ligne n°459 : series of possible translations in the target language. The next step ...
Ligne n°458 : ... corpus) creates a “translation model” that generates not one but a- Ligne n°459 : series of possible translations in the target language. The next step
Ligne n°460 : is running these possibilities through a monolingual language model in ...
Ligne n°464 : ... (Parallel human-translated corpora are hard to come by; large amounts- Ligne n°465 : of monolingual training data are not.) As with the translation model,
Ligne n°466 : the language model uses a brute-force statistical approach to learn ...
Ligne n°466 : ... the language model uses a brute-force statistical approach to learn- Ligne n°467 : from the training data, then ranks the outputs from the translation
Ligne n°468 : model in order of plausibility. ...- Ligne n°470 : Statistical machine translation rekindled optimism in the field.
Ligne n°471 : Internet users quickly discovered that Google Translate was far better ...
Ligne n°475 : ... result. And language pairs like Chinese-English, which are unrelated- Ligne n°476 : and structurally quite different, make accurate translation harder than
Ligne n°477 : pairs of related languages like English and German. But more often than ...- Ligne n°500 : Neural-network-based translation actually uses two networks. One is an
Ligne n°501 : encoder. Each word of an input sentence is converted into a ...
Ligne n°505 : ... a private research organisation, uses an intriguing analogy to compare- Ligne n°506 : neural-net translation with the phrase-based kind. The latter, he says,
Ligne n°507 : is like describing Coca-Cola in terms of sugar, water, caffeine and ...
Ligne n°511 : ... Once the source sentence is encoded, a decoder network generates a- Ligne n°512 : word-for-word translation, once again taking account of the immediately
Ligne n°513 : preceding word. This can cause problems when the meaning of words such ...- Ligne n°518 : Neural-network translation requires heavy-duty computing power, both
Ligne n°519 : for the original training of the system and in use. The heart of such a ...
Ligne n°521 : ... or specialised hardware like Google’s Tensor Processing Units (TPUs).- Ligne n°522 : Smaller translation companies and researchers usually rent this kind of
Ligne n°523 : processing power in the cloud. But the data sets used in neural-network ...- Ligne n°528 : Fully automated, high-quality machine translation is still a long way
Ligne n°529 : off. For now, several problems remain. All current machine translations ...
Ligne n°528 : ... Fully automated, high-quality machine translation is still a long way- Ligne n°529 : off. For now, several problems remain. All current machine translations
Ligne n°530 : proceed sentence by sentence. If the translation of such a sentence ...
Ligne n°529 : ... off. For now, several problems remain. All current machine translations- Ligne n°530 : proceed sentence by sentence. If the translation of such a sentence
Ligne n°531 : depends on the meaning of earlier ones, automated systems will make ...
Ligne n°540 : ... resources are thin on the ground. For example, there are few Greek-Urdu- Ligne n°541 : parallel texts available on which to train a translation engine. So a
Ligne n°542 : system that claims to offer such translation is in fact usually running ...
Ligne n°541 : ... parallel texts available on which to train a translation engine. So a- Ligne n°542 : system that claims to offer such translation is in fact usually running
Ligne n°543 : it through a bridging language, nearly always English. That involves ...
Ligne n°543 : ... it through a bridging language, nearly always English. That involves- Ligne n°544 : two translations rather than one, multiplying the chance of errors.
- Ligne n°546 : Even if machine translation is not yet perfect, technology can already
Ligne n°547 : help humans translate much more quickly and accurately. “Translation ...
Ligne n°559 : ... anyone with parallel corpora to hand. A specialist in medical- Ligne n°560 : translation, for instance, can train the system on medical translations
- Ligne n°560 : translation, for instance, can train the system on medical translations
Ligne n°561 : only, which makes them far more accurate. ...
Ligne n°564 : ... optimised for the shorter and simpler language people use in speech to- Ligne n°565 : spew out rough but near-instantaneous speech-to-speech translations.
Ligne n°566 : This is what Microsoft’s Skype Translator does. Its quality is improved ...
Ligne n°572 : ... software allowing companies quickly to combine the best of MT,- Ligne n°573 : translation memory, customisation by the individual translator and so
Ligne n°574 : on. Translation-management software aims to cut out the agencies that ...
Ligne n°576 : ... translators. Jack Welde, the founder of Smartling, an industry- Ligne n°577 : favourite, says that in future translation customers will choose how
Ligne n°578 : much human intervention is needed for a translation. A quick automated ...
Ligne n°577 : ... favourite, says that in future translation customers will choose how- Ligne n°578 : much human intervention is needed for a translation. A quick automated
Ligne n°579 : one will do for low-stakes content with a short life, but the most ...
Ligne n°588 : ... small but much-admired startup, Lilt, uses phrase-based MT as the basis- Ligne n°589 : for a translation, but an easy-to-use interface allows the translator
Ligne n°590 : to correct and improve the MT system’s output. Every time this is done, ...
Ligne n°590 : ... to correct and improve the MT system’s output. Every time this is done,- Ligne n°591 : the corrections are fed back into the translation engine, which learns
Ligne n°592 : and improves in real time. Users can build several different memories—a ...
Ligne n°593 : ... medical one, a financial one and so on—which will help with future- Ligne n°594 : translations in that specialist field.
Ligne n°596 : ... TAUS, an industry group, recently issued a report on the state of the- Ligne n°597 : translation industry saying that “in the past few years the translation
- Ligne n°597 : translation industry saying that “in the past few years the translation
Ligne n°598 : industry has burst with new tools, platforms and solutions.” Last year ...
Ligne n°601 : ... the quality of MT will keep improving, and that for many applications- Ligne n°602 : less-than-perfect translation will be good enough.
Ligne n°603 : The “translator” of the future is likely to be more like a ...
Ligne n°734 : ... Google points out why. Automated speech recognition and machine- Ligne n°735 : translation have something in common: there are huge stores of data
Ligne n°736 : (recordings and transcripts for speech recognition, parallel corpora ...
Ligne n°736 : ... (recordings and transcripts for speech recognition, parallel corpora- Ligne n°737 : for translation) that can be used to train machines. But there are no
Ligne n°738 : training data for common sense. ...