News Categorization using Hybrid BiLSTM-ANN Model with Feature Engineering
Sowmya Sanagavarapu, Sashank Sridhar, S. Chitrakala · 2021
There is an exponential increase in the number of articles across diverse topics published in various media around the globe. A digital management system would help users to find the relevant topics of discussion in the system and help to find articles of their interest. This project implements a categorization model that uses a hybrid model consisting of BiLSTM and ANN that performs the classification of news articles into chosen topics by using the hypernyms and hyponyms of the words present in them. The BiLSTM model maps the semantic structure of the articles and the ANN model maps the semantic meaning of the article with the help of hypernyms and hyponyms. The performance of the hybrid classification model for news articles spread across six categories is measured over different chosen levels of hypernyms and hyponyms was analyzed. The model performed the best when the hybrid combination of Level 1-2-3 of hypernyms along with the words of the articles is used, achieving a high testing accuracy of 61.33%.