An Optimization Based Feature Extraction And Machine Learning Techniques For Named Entity Recognition In Biomedical Field

Abstract

Unstructured and structured documents processing are involved in Named Entity Recognition (NER) for recognizing definite entity classes and categorization of these entities into some predefined classes. Biomedical instances like RNAs, DNAs, disorders, viruses, proteins, genes and chemical components are identified using Biomedical Named Entity Recognition (BNER). Techniques used to extract those entities plays a major role in this BNER.Supervised Machine Learning (SML) approaches are used in various BNER techniques. In these approaches, in order to enhance the recognition process’s effectiveness, these features are used. A set of distinguishing and discriminating characteristics are used for identifying features, which is having an ability for indicating entity occurrence.Biocurators annotates only limited number of articles also consumes more processing time.In this work, propose an Enhanced System for Curatable-Biomedical Named Entities Recognition (ECBNER) and feature extraction approaches for bio-medical named entity recognition using aImproved Particle Swarm Optimization (IPSO). In machine learning, fusion of classifiers gives a fruitful research direction and it is an effective technique of it. A standalone classifier’s performance in classification can be enhanced using this. Result of various classifiers combination are aggregated to overcome individual classifier’s possible local weakness for producing highly robust recognition.Gene/Disease NER are processed under Conditional Random Field (CRF) and all action terms are collected and processed in concurrent manner to extract the accurate biomedical named entities.Finally this general framework to learn the representation by combining general and domain-specific features is proposed and evaluated, showing empirical results compare to the existing frameworks.