Google Neural Machine Translation (GNMT) is a neural machine translation system (NMT) developed by Google and introduced in November 2016, which uses artificial neural networks to improve fluency and accuracy in Google Translate.
GNMT improves the quality of translation by implementing an example-based machine translation method (EBMT) in which the system "learns from millions of examples". The architecture of the learning system proposed by GNMT was first tested in over a hundred languages ​​supported by Google Translate. With a large end-to-end frame, the system learns from time to time to create better and more natural translations. GNMT is able to translate whole sentences at once, not just piece by piece. The GNMT network can perform translation of interlingual machines by encoding sentence semantics, rather than by memorizing phrase-to-phrase translation.
Video Google Neural Machine Translation
History
The Google Brain project was established in 2011 in the "secret Google lab research lab" by Google Associates Jeff Dean, Google Researcher Greg Corrado, and Stanford University Computer Science professor Andrew Ng. Ng's work has resulted in some of the biggest breakthroughs on Google and Stanford.
In September 2016, a team of researchers at Google announced the development of the Google Neural Machine Translation (GNMT) system and in November Google Translate began using nerve machine translation (NMT) in preference to previous statistical methods (SMT) that have been in use since October 2007, with technology Own internal SMT.
The Google Translate NMT system uses a large artificial neural network capable of deep learning. Using millions of examples, GNMT improves the quality of translation, using the broader context to deduce the most relevant translations. The results are then rearranged and adapted to the grammatical approach of human language. The architecture of the learning system proposed by GNMT was first tested in over a hundred languages ​​supported by Google Translate. GNMT does not create its own universal interlingua but rather is directed to the similarities found among many languages, which are considered more attractive to psychologists and linguists than to computer scientists. The newly first translation engine is enabled for eight languages: to and from English and French, German, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish by 2016. In March 2017, three additional languages ​​are enabled: Russian, Hindi , and Vietnam along with Thailand whose support was added later. Support for Hebrew and Arabic is also added with help from Google Translate Community in the same month. In mid-April 2017 Google Netherlands announced support for Dutch and other European languages ​​related to English. Further support is added for nine Indian languages, ie. Hindi, Bengali, Marathi, Gujarati, Punjabi, Tamil, Telugu, Malayalam and Kannada at the end of April 2017.
Maps Google Neural Machine Translation
Languages ​​Supported by GNMT
This is a list of language translation pairs supported by Google Translate Engine Neural Machine Translation (NMT). Starting July 2017 all languages ​​currently only support translations to and from English:
Zero-shot Translation
The GNMT system is said to represent an improvement over the previous Google Translation because it can handle "zero-shot translations," which directly translate one language to another (for example, Japanese to Korean). Google Translate first translated the source language into English first and then translated English into the target language rather than translating directly from one language to another.
See also
References
External links
- Machine Translation Gains and Drawbacks
- Machine Translation Statistics
- International Association for Machine Translation (IAMT)
- Machine Translation Archive by John Hutchins. Electronic repository (and bibliography) of articles, books and papers in the field of machine translation and computer-based translation technology
- Machine translation (computer-based translation) - Publications by John Hutchins (including PDFs from several books on machine translation)
Source of the article : Wikipedia
0 Comments