Liuliu writeup of Borthwick et al.

This is a review of Borthwick_1998_exploiting_diverse_knowledge_sources_via_maximum_entropy_in_named_entity_recognition by user:Liuliu.

This paper is about using maximum entropy model to do name entity recognition.

I like

• The use of maximum entropy model in solving this problem was new at that time.

• They had a comprehensive set of features.

• The use of maximum entropy model kept people from writing patterns by hand.

I don't like

• Although M.E. could automatically learn the weights of each feature once features are selected, how to select the original set of features is not trival.
  • In fact feature selection is most of what the paper is about!