3 ms·
I used the Stanford nlp in a project last year, apparently it can (with appropriate models) handle German, Chinese (not sure which flavour) and Japanese (I thin
by stephenr 11y ago
I used the Stanford nlp in a project last year, apparently it can (with appropriate models) handle German, Chinese (not sure which flavour) and Japanese (I think?).
Unfortunately what I needed to process was likely to be in either English or Indonesian, without forewarning which it was, so we had to use an api (we used Google in the end , but the Microsoft one worked too) to detect + translate the text to English, which has some interesting side affects when dealing with numbers with suffixes, and dates, due to limitatioms of what Stanford nlp recognised (1 million is recognised as 1000000, 1M is not)
Also, handling informal text can be a nightmare - I'm still wondering if a mechanical Turk solution wouldn't have been better.