Note: this tagger has been replaced by the better-performing LBJ NER tagger, which is also available from the group software page.
The Named Entity Tagger is a self-contained package which incorporates versions of SNoW and FEX, together with an inference module and a gazetteer-based NE tagger. It includes a network trained to recognize Person, Location, Organization and Misc. entities in English.
The tagger uses SNoW (a network of classifiers) to classify individual words as Named Entity boundaries, and the inference module determines the most likely allocation of labels subject to some logical constraints (non-overlap, matching entity labels within boundaries). The result is a robust system capable of good performance on new data.
The most recent version offers command line options to control tokenization, and to allow access to the labels assigned by the gazetteer-based NE tagger.
