Nothing warms your heart quite like a bit of machine translated language that just doesn’t get it quite right. Sure there were examples of bad translation before (The term Chinglish/Chingrish is a testament to that) but the explosion of accessible Internet coupled with a few entrepreneurial types lead to this bad translation being commonplace anywhere multiple languages could be expected. So much so in fact that whole websites, which have been around for well over a decade, are now dedicated to the phenomena. Even the title of this blog post is a testament to the wonderfully horrible world of machine translation, thanks to this website.
So you can imagine the hilarity that ensued when Google announced it was going for a universal translator (much like the fictional Babel fish) that would leverage their translation engine as well as a good old fashioned text-to-speech program:
By building on existing technologies in voice recognition and automatic translation, Google hopes to have a basic system ready within a couple of years. If it works, it could eventually transform communication among speakers of the world’s 6,000-plus languages.
The company has already created an automatic system for translating text on computers, which is being honed by scanning millions of multi-lingual websites and documents. So far it covers 52 languages, adding Haitian Creole last week.
Don’t get me wrong though the idea intrigues me quite a lot. I spent the better part of 6 years learning Japanese and was still a while off from being fluent in the language. Sure I knew enough to get by in Japan (and really 2 weeks in the country taught me more than the previous 3 years did) but I’d have to also put that down to the majority of Japan’s younger generation knowing a substantial amount of English. Being able to connect with people who I don’t share a common language with is definitely something I can appreciate.
Whilst I don’t doubt Google’s ability to actually bring about a system like this, in fact I believe that you could probably cobble together a system right now using a couple discrete products, there’s some fundamental aspects to languages that will make the use less than optimal for many situations. For example let’s take a basic sentence in English and compare it to Japanese equivalent:
- We went to the pool.
- 私たちはプールに行きました (to be fair I verified that using Google translator, and it was spot on)
Now these are both simple sentences but the differences are in the construction. If we take the Japanese sentence literally it actually says “We pool went” which whilst understandable isn’t exactly English. This is because, in the most basic sense, Japanese sentences are constructed in the form of <subject><object><verb> whereas English comes in the form of <subject><verb><object> (an in depth analysis of these two grammatical structures is shown here and here respectively). For text translation this doesn’t matter too much since you can program your algorithm to detect the different grammatical structures.
For real time translation however it becomes more difficult. Not only do you have to detect the appropriate sentence structure you also have to figure out when someone has finished a sentence. Otherwise if you’re just translating word by word your results will end up like I showed above, with sentences making sense only after you rework them in your head. Additionally more complicated sentences can change the meaning of certain words so that a blow by blow translation ends up not making any sense at all. There is of course one way to get around this and the translation service not real time but on a per sentence basis. That however is not what has been trumped up in the media over the past couple days.
In fact devices that can do a sentence by sentence translation (with the appropriately Stephen Hawking-esque voice) already exist, especially in Japan. I was lucky enough to be able to borrow one such device from my teacher whilst staying in Japan and many of the people we ran into had one as well. Granted their translation capabilities were limited when compared to that of Google but they were more than enough for the average tourist. I only wish I had picked one up whilst I was over there as the same devices cost about ten times as much over here.
I’ll be watching the development of this technology pretty closely, as I have done with Google Translate over the past few years. They’ve come a long way with developing their own translation engine and they have made significant leaps in the languages that were the source of so much ridicule in the past. As for the hopes of a real time universal translator I’d say there are some fundamental barriers to achieving it, but that won’t stop us from talking to each other with what will appear to be a satellite delay, even though we’re standing right in front of each other.
That would be rather amusing to watch, actually… 🙂