{"id":93513,"date":"2020-10-20T13:41:04","date_gmt":"2020-10-20T10:41:04","guid":{"rendered":"https:\/\/en.buradabiliyorum.com\/facebooks-new-ai-model-translates-100-languages-without-going-through-english-first\/"},"modified":"2020-10-20T13:41:04","modified_gmt":"2020-10-20T10:41:04","slug":"facebooks-new-ai-model-translates-100-languages-without-going-through-english-first","status":"publish","type":"post","link":"https:\/\/buradabiliyorum.com\/en\/facebooks-new-ai-model-translates-100-languages-without-going-through-english-first\/","title":{"rendered":"#Facebook\u2019s new AI model translates 100 languages \u2014 without going through English first"},"content":{"rendered":"

“#Facebook<\/a>\u2019s new AI model translates 100 languages \u2014 without going through English first<\/strong>”
\n<\/p>\n

\n Facebook has\u00a0open-sourced an AI model<\/a> that can translate between any pair of 100 languages without first translating them to English as an intermediary step.<\/p>\n

The system, called M2M-100, is currently only a research project, but could eventually be used to translate posts for Facebook users,\u00a0nearly two-thirds of whom use a language other than English.<\/p>\n

\u201cFor years, AI researchers have been working toward building a single universal model that can understand all languages across different tasks,\u201d said Facebook research assistant Angela Fan in a blogpost<\/a>. <\/span><\/p>\n

\u201cA single model that supports all languages, dialects, and modalities will help us better serve more people, keep translations up to date, and create new experiences for billions of people equally. This work brings us closer to this goal.\u201d<\/span><\/p>\n

[Read:\u00a0Researchers use AI to translate text found on ancient clay tablets]<\/em><\/p>\n

The model was trained on a dataset of 7.5 billion sentence pairs across 100 languages that were mined from the web. Facebook says all of these resources are open source and use publicly available data.<\/p>\n

To manage the scale of the mining, the researchers focused on language translations that were most commonly requested and avoided the rarer ones, such as Sinhala-Javanese.<\/p>\n

They then grouped the languages into 14 different groups, based on linguistic, geographic, and cultural similarities. This app<\/a>roach was chosen because people in countries with languages that share these characteristics would be more likely to benefit from translations between them.<\/p>\n

For instance, one group included common languages in India, such as Hindi, Bengali, and Marathi. All the possible language pairs within each group were then mined.<\/p>\n