Statistical Approaches to Biomedical Entities Recognition

Abstract

Named Entity Recognition (NER) is one of essential tasks for knowledge acquisition from scientific literature. In this paper, a full automatic named entities recognition from biomedical literature is presented by using Hidden Markov Model in which a rich set of features are concerned and back-off strategy is employed to overcome data sparseness problem. Experiments with GENIA corpora of different versions showed that the presented approach achieved promising results of 76% and 62% F-score for singular-type and multiple-type entities recognition respectively.