Viswanathan K's Thoughts

It’s possible to translate stuff from one language to another without having any clue about either language involved! After all, that’s how Google Translate operates: it’s just an algorithm (albeit an extremely good one) that mines a huge amount of articles in different languages to create its word/phrase map. Which it then uses to translate. No human in involved in translating anything. Is it perfect? No. Is it good enough for most purposes? Absolutely, yes. I remember this article a long time back by Kevin Kelly on whether this “Google approach” could be applied to science as well: have an algorithm look at all the data we get from all the sensors, satellite pics and Large Hadron Colliders of the world. If the algorithm is smart enough, it’ll come up with patterns of nature and even make predictions. Of course, that approach doesn’t lead to any understanding or comprehension, but do we even need to worry about understanding, Kelly asked? I always felt the answer to that question w...

It seems to be the general trend in everything. The microprocessor-software combo seems to do most things that was done by humans much faster and way better. Except language. That’s one area that still resists the tech juggernaut. Items related to language, like translations and speech, seem too tough to crack. Let’s start with translation. Word Lens , an app developed for the iPhone, became very famous due to videos like this: Point your phone, take a picture and voila! The translated text appears. Awesome, isn’t it (even if it only worked between English and Spanish)? People certainly thought so and shelled out money to buy the app. Which is when the complaints started: the translations weren’t always right. Nor were they context sensitive. Google Translate creates and scans (massive) databases, and then applies algorithms to decide (guess?) which words mean the same thing in different languages. Now Google’s taken it to the next level by combing translations with speech recognition:...