The art of domain classification and recognition for text conversation using support vector classifier

Abstract

This paper presents an art for recognition of text conversation into multiple domain categories using SVC. The whole process of recognition includes diverse components as: lexical analysis, feature extraction, features normalisation, feature reduction and finally recognition. Useful words are extracted using lexical analysis from the input text paragraph and transformed it into a binary matrix for further processing through feature extraction. Feature normalisation is used to normalise the values of binary matrices while feature reduction is done through principal component analysis to extract the important features from the feature vector and now it is passed to different configurations of SVM, then the best one is selected for the final process of classification and recognition. The domain's categories are defined on the basis of various real life situations and conversation to train the system like education and research, personal, patriotism, terrorism, medical, religious, sports, and business. The experimental results demonstrate that the proposed approach works effectively with about 75% accuracy.