Sorry for my unfamiliarity with the field. Hopefully this question is on-topic here.

Currently the machine translation from western European languages to English is arguably quite robust, with Google Translate able to quite accurately translate articles on Wikipedia or news site. However, it seems to me that the translation between Chinese/Japanese and English is still quite off: Google Translate is frequently unable to produce a reasonably coherent/meaningful translation for even just one Chinese/Japanese sentence.

It gets me wondering, is the state-of-the-art in CJK machine translation still much off the pace compared with European languages? Or is it just that some research advances haven't been applied to the industry yet? What is the current status of the cutting-edge researches on this field, e.g. novel translation models/improvements on precision rate. Where should I go to look for more information?

  • 1
    $\begingroup$ A reference request like yours is too broad for Stack Exchange -- you ask for a survey of a whole research area! You need to narrow your focus considerably before a question of reasonable scope appears. Try talking to your advisor(s), search with Google Scholar and check out this guide to better (re)searches on Academia. $\endgroup$ – Raphael Feb 26 '16 at 14:25
  • $\begingroup$ FWIW, there is also Linguistics where some NLP and computational linguistics folks hang out. But don't repost this exact question there; it's too broad no matter the site. You may want to visit their chat, though. $\endgroup$ – Raphael Feb 26 '16 at 14:26
  • $\begingroup$ Nota bene: translations by Google between English and German are horrible as well. The ngram approach (which is afaik still the state-of-the-art, but I'm not an expert) has obvious limits w.r.t. long side clauses or references, and nested structures. $\endgroup$ – Raphael Feb 26 '16 at 14:28
  • $\begingroup$ That this question is closed is disappointing. :( CJK machine translation is indeed pretty hard to search for, and when searching this was the first promising thing I came across. It's true that the vast majority of published research on machine translation is Eurocentric. Like @Raphael said, I think techniques haven't been adapted very well for significantly different word order -- which CJK certainly have. $\endgroup$ – Alex Meiburg Jul 10 '17 at 2:31