Enhancing Learning Retention with Document Summarization Using NMF and ACO

Abstract

Retaining learning is a desirable value for the learned. The traditional approach to learning retention is to read, understand and summarise. This explains the reason why vital points in voluminous book are highlighted for the purpose of retention. This study developed an unsupervised extractive summarization algorithm using Non-negative Matrix Factorisation and Ant-Colony Optimization Techniques for electronic text summarization. The developed system is an improvement over the existing NMF and LSA in the sense that LSA and NMF do not adequately address the noise issue that features in semantic document representation thereby leading to poor selection of meaningful sentence that represent the document summary. While NMF was used for factorising the initial Document-Term matrix generated from the document, ACO was used to remove the remnant of noise from the DTM. The improved algorithm is applied on a literature text material which reduces the voluminous material into handy summary by extraction. The algorithm was evaluated using compression ratio, retention ratio and was found to be adequate. Keywords: Text Summarization, Non-Negative Matrix Factorisation, Ant-Colony Optimization