I recently interviewed Microsoft's Garry Wiseman. Garry is the group program manager for the commerce search team in the Live Search group. We spoke at length about Microsoft's new Shopping Search engine. I covered the announcement of the new shopping search in detail previously on Search Engine Watch, and I wanted to get under the covers a little bit to learn more about how it worked.
One of the most intriguing things that Microsoft did with this release is start publishing detailed ratings for a large number of characteristics for consumer electronics devices. You can get to these by searching on a consumer electronics device, for example, iPod Nano, and then clicking on one of the 4 pictures presented in the results to get to the product details page. For example, you might see ratings for up to 40 different characteristic of the device. For an iPod Nano, 15 characteristics including price, sound quality, size, ease of use, battery life, screen, memory, instructions, and more.
According to Garry, this information is extracted from the web using a "sentiment extraction engine". This engine is used to automatically scan trusted review sites and then find out what are the attributes of a product that users comment on in reviews. Then, the engine looks at what people say about each attribute, to determine whether the review is positive or negative for that given attribute. This data is then all tabulated and presented on the product details page for that product.
Garry was not able to comment on whether or not this sentiment extraction engine might be used in other verticals in the future. Clearly, this depends on their being enough data available across the web for them to extract meaningful information. I suggested to Garry that they consider implementing a master product directory for consumer electronics to make the navigation from product a little easier than using search as the interface to all this great information. Understandably, he could not comment on that either.