Trained Named Entity Recognition Using Distributional Clusters

Abstract

This work applies boosted wrapper induction (BWI), a machine learning algorithm for information extraction from semi-structured documents, to the problem of named entity recognition. The default feature set of BWI is augmented with features based on distributional term clusters induced from a large unlabeled text corpus. Using no traditional linguistic resources, such as syntactic tags or special purpose gazetteers, this approach yields results near the state of the art in the MUC 6 named entity domain. Supervised learning using features derived through unsupervised corpus analysis may be regarded as an alternative to bootstrapping methods.