Selen writeup of Krishnan 2006

This is a review of Krishnan_2006_an_effective_two_stage_model_for_exploiting_non_local_dependencies_in_named_entity_recognition by user:Selen

Things I like about this paper:

They capture the non local dependencies in a named entity extraction tasks eliminating the need of doing approximate inference hence their method is efficient

Their method is generalizable unlike Sutton 2004 paper which was limited to basic entities such as capitalized words

Things I don't like about this paper:

They compare their performance to baseline CRFs but they don't compare it to other methods' that are developed to capture non-local dependencies.