Fiete - 5:56 pm on May 20, 2003 (gmt 0)

Probably FAST uses for the clustering a lot of different available information sources.

I'm quite certain that they use ODP.

If the cluster engine finds in the results several pages that are listed in - and linked from - an ODP category, it is very likely that this information is used for the cluster. That does not necessarily mean that they use the ODP taxanomy. It may get into the calculation indirectly, but it's also likely, that they use it not as the only method.

Btw, I think I have heard that they use also some semantic logic in order to build their clusters.