Supervised Machine Learning Approach for Bio-molecular Event Extraction

Abstract

The main goal of biomedical text mining is to capture biomedical phenomena from textual data by extracting relevant entities, information and relations between biomedical entities such as proteins and genes. Most of the research in the related areas were focused on extracting only binary relations. In a recent past, the focus is shifted towards extracting more complex relations in the form of bio-molecular events that may include several entities or other relations. In this paper we propose a supervised approach that enables extraction, i.e. identification and classification of relatively complex bio-molecular events. We approach this as the supervised machine learning problems and use the well-known statistical algorithm, namely Conditional Random Field (CRF) that makes use of statistical and linguistic features that represent various morphological, syntactic and contextual information of the candidate bio-molecular trigger words. Firstly, we consider the problem of event identification and classification as a two-step process, first step of which deals with the event identification task and the second step classifies these identified events to one of the nine predefined classes. Thereafter, we perform event identification and classification together. Three-fold cross validation experiments on the Biomedical Natural Language Processing (BioNLP) 2009 shared task datasets yield the overall average recall, precision and F-measure values of 58.88%, 74.53% and 65.79%, respectively, for the event identification. We observed the overall classification accuracy of 59.34%. Evaluation results of the proposed approach when identification and classification are performed together showed the overall recall, precision and F-measure values of 59.92%, 54.25% and 56.94%, respectively.

All authors equally contributed for the paper.