Evaluating Extractive Summarization Techniques on News Articles

원문 초록

In recent years, due to the rise of deep learning and natural language processing, text summarization has become a huge topic among scholars. Text summarization derives a shorter coherent version of a longer document. There are two methods of summarization namely, abstractive and extractive. This paper focuses on extractive summarization using TextRank and BERT. These algorithms have been tested under various circumstances to determine the best and they all perform better on certain parameters. The goal of this paper is to determine which algorithm performs better as compared to human generated extractive summaries on news dataset. The same dataset was used for both these algorithms and the summaries were evaluated using ROUGE Score. The result showed that TextRank yielded a better ROUGE score as compared to BERT. TextRank showed higher F-measure and recall while BERT had higher precision.