Optimized Version of Tree based Support Vector Machine for Named Entity Recognition in Medical Literature

Abstract

Medical literature comprises valued information, like clinical symptoms, diagnosis, dosage, and treatment for specific diseases of a particular disease. Named Entity Recognition (NER) is a primary process involved in the extraction of knowledge from unstructured text and providing it as a Knowledge Graph (KG). Several existing works of NER suffers from small scale human- labeled training dataset. Since the extraction of knowledge from medical literature is a difficult process, this paper aims to focus on the development of the NER model using machine learning (ML) approaches to improve efficiency. The proposed model is based on the optimized version of support vector machine (O-SVM) where the optimal parameters of the tree based SVM are tuned by the particle swarm optimization (PSO) algorithm. In addition, the medical dataset is initially preprocessed and then the classification process takes place via the O-SVM model. The weight and bias parameters in tree based SVM model are tuned by the PSO algorithm. The experimental results analysis of the O-SVM model is carried out and the results are compared with the state of art approaches in terms of diverse measures.