The Improvement of Biomedical Named Entity Recognition with Semi-joint labeling presentation

Abstract

Named entity recognition (NER) is an important step for the information retrieval. In biomedical field, due to the fact that there is no community-wide agreement on how a particular name entities (NEs) should be named. To improve the performance of NER in biomedical field, this paper proposed to use a new tagging presentation, semi-joint labeling. This presentation works by adding chunking information on the tags which are not NEs. With the proposed presentation, we improve the F-score from 62.37% with IOB2 tagging presentation to 63.33%, and right and left boundary matches improves 0.88% and 0.56% respectively.