Applying the Multiple Cause Mixture Model to Text Categorization

Machine Learning in Information Access

Volume

Papers from the 1996 AAAI Spring Symposium

Contents

This paper introduces the use of the Multiple Cause Mixture Model to automatic text category assignment. Although much research has been done on text categorization, this algorithm is novel in that is unsupervised, that is, does not require pre-labeled training examples, and it can assign multiple category labels to documents. In this paper we present very preliminary results of the application of this model to a standard test collection, evaluating it in supervised mode in order to facilitate comparison with other methods, and showing initial results of its use in unsupervised mode.