Steve Rubel makes some interesting suggestions about the need for better tools to discover blogs on certain topics. I agree that this is something which has yet to be cracked in any comprehensive way (though Sphere is possibly ahead of the game in terms of thinking about this one). However, I don't agree with Steve's comment that Amazon has done a good job of concordance creation for books, at least not by judging the example that he points to. The example is for the book 'Quicksilver' by Neal Stephenson and contains approximately 100 terms. The fact that this 100 includes morphological variations on general verbs (come, came, go, going, went, look, looked) suggests that they could be providing far more interesting information about the book.
A better candidate feature from Amazon may be their SIPs (statistically improbable phrases) which do a great job for All the President's Men: bugging operation, political espionage, secret fund, finance chairman, though not such a good one for Stephenson's book. The key difference is one needs to discover terms and phrases which do the best job of describing the content of the text as well as differentiating it from other texts. Terms like 'go', 'come', 'look' aren't going to help you there.
Comments