Summize – Search Engine for Reviews

Just ran across Summize while catching up on the Lifehacker syndication feed. The main feature Summize provides is an aggregation of reviews of books, movies, videos and other products from blogs and ecommerce sites like Amazon. But the site is much deeper than that.

Recently Amazon started showing the breakdown of the absolute counts for each level of a 1 to 5 star product rating. This sometimes provides far more insight than you get from a single averaged value, especially when there are just a few reviews. Summize provides a similar breakdown via a color bar (red = terrible, green = great). When there are a small total number of reviews for a product, just a few poor ratings can have a significant effect on the overall rating. If you then look at the actual reviews, you might discover that the highly negative reviewers bought the wrong product and are mad, hated a feature you don’t care about, or are simply stupid. Same goes for reviewers giving very high reviews because they love a feature you could care less about.

Summize also provides a list of links to related items. For example, last night I watched MirrorMask, which was written by one of my favorite authors, Neil Gaiman. Searching for MirrorMask brought up reviews of not just the movie, but also the illustrated film script, a book on the making of the movie, the soundtrack and more. Only one item was unrelated to the movie.

Links to other topics are also provided. For MirrorMask, links were provided to pages for Neil Gaiman and Dave McKean, as well as to some of the actors.

Although Amazon and many other sites conveniently provide ratings on a 1 to 5 scale or on a different scale that can be easily scaled to 1 to 5, obviously most bloggers don’t provide numerical ratings in an easily retrievable format. It appears that Summize has developed an algorithm for parsing the text of the review and using sentiments expressed in the review (fantastic, sucked, godlike) to convert the review to a 1 to 5 rating. I’m assuming they are doing more than looking for adjectives to which they have assigned values, but, hey, even that might work pretty well in aggregate.

So, how do they make money? Product links on Amazon, for starters.

Obviously, they are pulling in a lot of data and have one some interesting stuff with it. I recommend digging around on the site on the trends page and elsewhere to see what they have done with the data so far.