[Topic-models] Clustering per-document topic vectors

Fisher, Douglas H. dhfisher at nsf.gov
Sun Nov 9 18:21:29 EST 2008


Does anyone have experience or other observations on clustering
per-document topic vectors produced by application of topic modeling to
a document collection, particularly hard-assignment, hierarchical
clustering (e.g., producing a dendogram), from which an analyst might
more easily assess various partitions of the documents (e.g., so that an
analyst can cut the dendogram to give a three-cluster partition of the
documents and compare this to four- and five- cluster partitions
produced by different cuts)? Best, Doug
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/topic-models/attachments/20081109/1532d331/attachment.html>


More information about the Topic-models mailing list