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

Similar things have been around for a long time.

For instance, you can train something like LDA on the contexts that words appear in. Also I've seen a system that uses a neural net to generate a 16,000-bit vector and very interesting things happen when you use AND and OR.



Agreed. I once used a rich vector space model to answer natural language questions related to geography with a similar approach.




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

Search: