A statistical technique for bootstrapping available resources for proper nouns classification

Describes an algorithm for improving the performance of unknown proper noun recognizers, using a statistical framework. We present a bootstrapping technique that starts out by using a training set to acquire contextual classification cues, and then uses the results of the initial phase to acquire additional training data from an unlabeled corpus. The training set (tagged proper nouns in contexts) is obtained trough an application of standard knowledge-based techniques for proper noun tagging, commonly used in information extraction systems.