Using weighted majority voting classifier combination for relation classification in biomedical texts

• Published in: 2014 International Conference on Control, Instrumentation, Communication and Computational Technologies (ICCICCT) pp. 1205 - 1209
• Main Authors: Remya, K. R., Ramya, J. S.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-07-2014
• Subjects: Accuracy; biomedical text classification; Classification algorithms; Data mining; Diseases; Instruments; Kernel; Machine learning; multiple classifier systems; relation classification; Training; weighted majority voting
• ISBN: 9781479941919; 1479941913
• DOI: 10.1109/ICCICCT.2014.6993144

In biomedical area, information is mainly in natural language text format. Such information is stored in huge repositories. It is not easy to access required information from this large amount of data. Also the classification systems developed for general text is not applicable for biomedical data. The biomedical researchers need fast and accurate information accessing tools for extracting useful information from huge amount of biomedical repositories. This paper proposes a multiple classifier system for relation extraction from biomedical sentences. For classifier combination, the system uses weighted majority voting method. It classifies biomedical sentences according to the disease-treatment relations present in the sentences. This paper shows that multiple classifier system outperforms the single classifiers for relation classification from biomedical sentences.