tedster - 9:37 pm on Feb 26, 2010 (gmt 0)
I can understand why they chose the Reviews to be at the top of the list for testing.
Me too. I've written before about what a huge challenge it is to do REAL sentiment analysis. But if review information is all nicely packaged up, then you don't need full blown SA.
I've even read recent articles that call review data "sentiment analysis". What garbage! There's no analysis involved at all, just data retrieval. You might as well call a visual inspection of a TOC "information retrieval".