Well, it's not that easy. The algorithms are very primitive and too full of noise to be useful. For example, try this on restaurant reviews like http://www.yel

Well, it's not that easy. The algorithms are very primitive and too full of noise to be useful. For example, try this on restaurant reviews like http://www.yelp.com/biz/el-gaucho-seattle http://www.yelp.com/biz/el-gaucho-seattle. I get these tags: steak reviews seattle food service gaucho restaurant review Not useful, right? The current state of the art would use much more sophisticated NLP for generating POS tags and use sentiment analysis. For example, check out MSR Splat at http://research.microsoft.com/en-us/projects/msrsplat/default.aspx http://research.microsoft.com/en-us/projects/msrsplat/defaul....