A Case for Human Translation
By Darron | machine translation | No comment yet
I’m all for machine translation. I studied it in school and worked on an inter-university project 10 years ago whose goal was to incorporate it into AI systems. I’ve also looked several times into the feasibility of leveraging the technology more succesfully at my current company. In each case I’ve come to the same result; the state of the art is insufficient to rely on for quality, lasting translation…at least in many areas. Some technical language (or United Nations speeches) come out great. Other areas like advertising and literature get pretty garbled. Case in point, the Chinese restaurant pictured above which is undoubtedly not named ‘Translate Server Error’ in Chinese.
Google Translate and Global eLearning
By Darron | Emerging Technology | No comment yet
In a recent article in the New York Times, Google Translate claims that using statistical machine translation (MT) it can translate text between 52 languages. According to co-founder Sergey Brin and principle scientist Franz Och there is still much work to be done in order to get the MT working so that the translations sound human. Achieving this is going to require amassing incredible amounts of data and then sorting that data in meaningful ways. Source of data must be considered, of course. Like IBM and Microsoft, Google has some linguistic corpora from United Nations proceedings. It’s clear advantage is the ability to glean more linguistic information from web users. Google also has computing power beyond the reach of most other enterprises if you lash together their network of data centers, which just makes processing the data that much faster. It’s only a matter of time before Google Translate is producing machine translations that sound human. And they’ll be able to do it in more than 52 languages with many different specialties (e.g. political speech, legalese, marketing speech). There is no telling when such a dream will be realized, but it is coming.
