Posted by Franz Josef OchMachine translation is hard. Natural languages are so complex and haveso many ambiguities and exceptions that teaching a computer totranslate between them turned out to be a much harder problem thanpeople thought when the field of machine translation was born over 50years ago. At Google Research, our approach is to have the machineslearn to translate by using learning algorithms on gigantic amounts ofmonolingual and translated data. Another knowledge source is usersuggestions. This approach allows us to constantly improve...
Monday, 29 September 2008
Subscribe to:
Posts (Atom)
