Monday, December 04, 2006

Automated translation (google it)

There have been translation tools that attempt to translate from one written language to another available for sometime. I remember even about eight years ago getting hold of a free version of some software that attempted to translate between Spanish and English. It was fairly useless but the output did provide some laughs for a Spanish speaker I was working with at the time.

Recent developments in the Google translation tools have impressed me somewhat more though.

Despite leaving it until almost the age of 40 before attempting to learn a second language, there are a number of moments in my life that I can identify as specially relevant to language learning.

I remember my father telling me a 'supposed true story' about a computer translation program in the 60's. The story goes that they developed a Computer program to translate from English to Russian. They fed the program the phrase "out of sight ,out of mind" and after some time it spat a result in Russian that was equivilent to "invisible, imbecile". My father knew even in the 70's when he told me that this was a joke in the armed forces (I don't know if they had a phrase for urban legends back then) and I even found a reference to it and similar versions on the web.

I believed him at the time (I was just a kid ;)) and even then it struck me how hard it would be to translate languages. "out of mind" is similar to "out of your mind" and can often be seen as "out of your mind with (worry, fear, anger etc)". The meaning of even that part of the phrase is hard to fathom unless you just know it.

Translation tools can be handy for short phrases, but tranlating either way is not reliable for learning purposes as the results are often really bad.

I still run the odd phrase through a translation tool every now and again (usually one I already have an expected answer for). I have notice recently that although still far from perfect the Google translation tools can be surprisingly accurate. Apparently Google is using their knowledge of the Web to 'brute force' the problem and is deriving translation information from the huge numbers of translated documents they have indexed.

A simple but illustrative example is given with the short phrase (wo3 hui4de) which I usually see in subtitles and Chinese media to mean something like "I will" or in some circumstances "I can", often provoked by someone else asking for something to be done. Google translates to "I will" :), babel fish translates to "my meeting" :(.

I am going to keep watching, but anything that is getting better is a good thing in my opinion.

It is worth looking at the translation page on the excellent MDBG dictionary though because although it uses babel fish it also break the Chinese you enter down in to word sized chunks, allowing you to make your own mess of the meaning.

1 comment:

Mutant Jedi said...

I use google language tools as well as a few other translation sites. The problem I have with them is that they don't really have any sense of context, grammar or shades of meaning. They're often really bad at translating 口語。While I like google for small chunks, I used it for a longer document, a friend's blog, and it became stupid gibberish. The reason for this is that she was using a descriptive metaphorical style. The tool didn't have a clue as to what she was trying to say.

I like using MacKEY5 (they also have a PC version). It will show you the pinyin and definitions of words as your mouse is over the characters. This helps me fill in any blanks I might have. (which is far more than I'd like)