Finding Structure in Noisy Text: Topic Classification and Unsupervised Clustering

In this paper we present recent advances in automatic categorization of noisy, unstructured text messages posted on newsgroups. Newsgroup messages pose a significant challenge to automatic text categorization due to broad coverage of subject matter and the use of informal language. This paper addresses (a) spotting messages that are on topics of interest to the user, and (b) automatic organization of a large corpus of messages without any prior knowledge about topics of interest to the user. We present supervised classification results using our hidden Markov model based topic classification engine on messages from two different newsgroup corpora. Given that in an operational settin