Ref “However, if you want to use this in a loop you are out of luck” - I think you can use a closure to get around this. My js is not top-notch so sorry if I’m wrong, but something like the following seems to work ok:

function requestOne(text, id) {
google.language.detect(text, function(result) {
if (!result.error) {
var language = ‘unknown’;
for (l in google.language.Languages) {
if (google.language.Languages[l] == result.language) {
language = l;
break;
}
}
var container = document.getElementById(id);
container.innerHTML = text + ” is: ” + language + “”;
}
});
}

for (var i = 0; i < texts.length; i++) {
this_text = texts[i];
this_id = ids[i];
requestOne( this_text, this_id );
}

Am I hallucinating?

Alex

It incorrectly identified Chinese as Italian, and Japanese as Slovak (I think; sk?) though. Which seems like pretty strange mistake to make, especially having heard that Google’s translator is quite good at translating Chinese (and Arabic) in particular. Then again, all that code means nothing to me, though (the language bit is what interested me) so it might be some limitation with how the API parses the text? Or that might not even make any sense. It’s a shame the demo page doesn’t output the confidence values too. I’d be interested to see that.

Another potential issue might be correctly identifying tweets in one language, that happen to contain words of another language. For example, it showed this tweet: “GIF Sundsvall vs. Helsingborgs IF (Allsvenskan): Match has finished. Result: 0:3″ (URL from tweet removed, lest I annoy your spam filters) as Swedish, despite obviously being English.

The code means very little to me, but the idea is definitely awesome.

It’s a shame automatic machine translation comes with so many inherent downfalls. Though I think Google’s is one of the better ones, with quite a shallow approach to disambiguation, which would be the biggest hurdle (if you even intended to advance it to actually translating the given text).