This talk will first describe the problem of noun phrase coreference, the task of determining which noun phrases in a text or dialogue refer to the same real-world entity. We will then explain the role of noun phrase co reference for one practical natural language processing application --- information extraction. Finally, we will introduce an algorithm for noun phrase co reference resolution that combines supervised and unsupervised learning techniques. The approach differs from existing methods in that it views co reference resolution as a clustering task. In initial evaluations on a standard coreference resolution corpus, our results are encouraging. The coreference clustering algorithm appears to provide a flexible mechanism for combining context-independent co reference constraints and context-dependent preferences for accurate partitioning of noun phrases into co reference equivalence classes.