Extracting Useful Review Snippets from Yelp Using Natural Language Processing

The code from the paper is available at http://groups.csail.mit.edu/rbg/code/content_structure/code....

You can see a demo of the site at condensr.com. There will be substantial changes to the site including predicting the best dish at a restaurant on the site soon. An academic paper should follow shortly.

Interesting fact: I was asked to program something that did just this for my interview with Yelp.

Anyone know what software (open source, Java/Python) that is out there that can give me similar results? I've looked at LingPipe, but the closest thing it does is sentiment classification.

Python's NLTK has a lot of the raw tools at hand to do NLP tasks.

Check out the open source Stanford Parser (Java) http://nlp.stanford.edu/software/lex-parser.shtml

I also wrote a free open source javascript API for the Stanford Parser if you want to play around with NLP in JS. http://nlp.naturalparsing.com/browserparser/parse (it is a bit slow at the moment, someone is using it to parse all of wikipedia).