Named Entity Recognition in Biomedical Literature Using Two-Layer Support Vector Machines

In this paper, we propose a named entity recognition system for biomedical literature using two-layer support vector machines. In addition, we employ a post-processing module called a boundary check module to eliminate some boundary errors, which can lead to improved system performance. Our system doesn't make use of any external lexical resources and hence it is a fairly simple system. Furthermore, with carefully designed features and introducing a second layer, our system can recognize named entities in biomedical literature with fairly high accuracy, which can achieve the precision of 83.5%, recall of 80.8% and balanced F-beta=1 score of 82.1%, an approximate state of the art performance for the moment.