Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I tried "the" first, and then "friend".

It looks like it's generated from statistics plus a language model plus the contents of the web (possibly and/or google books).

If that's the goal it looks like they're doing some great work. If being a dictionary is the goal it looks like there are currently better solutions.

However a) give it a few years and b) if they're able to extract that amount of information from their corpus, just imagine how that could be applied to a search engine - I get the feeling that they're showing their hand a little in their progress on the natural language processing front.

If that's so then I can't wait to see them apply it to search.



What you're seeing for English is the Collins Cobuild dictionary.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: