I've just been looking inside the (alpha release) of Boorah. The basics: a consumer facing review aggregation site which uses NLP to mine reviews and produce a score for each restaurant reviewed. Currently it is running over 95,948 reviews for restaurants in the bay area.
I like the format of the site: you can search for restaurant names, food types (Japanese, Indian, etc.) and by location. The results present a list of restaurants with their Boorah rating, the food type it has been classified into and a buzz score. For each restaurant there is a details pane which gives the name, address, etc. In addition there is a tabbed pane which provides a summary of reviews (snippets mined from the review text with sentiment/opinion text in it), reviews, reservations, maps and directions. In addition to the overall rating, there are ratings for food ambiance and service.
I haven't yet found much info about the technology. What I'm really wanting to know is what the ratings numbers are (both the evaluation numbers and the buzz scores). In addition, figuring out how they mine the opinion snippets would be particularly interesting. It is great to see more companies producing consumer facing social media platforms and portals which include sentiment mining of some form (see OpinMind for an earlier play in this area).
One of the problems with their rating is that if a place has low buzz, it gets a score of 0.0. Is that because there aren't enough reviews found, or because all the buzz is negative? I found a great example of a complex type of sentiment analysis in their snippets from this category of restaurants:
When we moved from the area, my wife's most frequent complaint was that no one made orange peel beef like Su Hong.
This is a positive review for the restaurant, but it mentions a complaint and so teasing apart the positive evaluation from the language is tricky.
Summary (after 15 mins on the site) : great to see some new stuff in this area, but some more information about their approach (and their company in general) would be useful for understanding what it is aiming to do and what the user's expectations should be.
Do you have any idea how much or what kind of NLP technology they are using? Sophisticated stuff could be useful, but I'm not sure if it would be worth the cost and complexity. Food is a good domain for it, tho. Tim
Posted by: tim finin | October 18, 2006 at 10:34 AM