Extractive Text Summarization Using LSTM-Based Encoder-Decoder Classification

Abstract

Nowadays, text summarization is one of the important areas to be focused on. As the World Wide Web is growing, a huge amount of text articles (especially blogs, scientific articles) are also generated on the internet. Automatic text summarization is one of the important techniques to shorten the original text in such a way that shorten or summarized text covers incisive and meaningful sentences of original huge text. Extractive summarization extracts important sentences from original documents and then aggregates all these sentences to generate the summary. We have proposed a novel LSTM based encoder-decoder, which plays a vital role in the extractive text summarization process. CNN news article dataset is utilized for training our model. Our model is evaluated on standard metrics like Gold Standard, Recall Oriented Understudy for Gisting Evaluation (ROUGHE)-1, and ROUGHE-2. After evaluation, our model achieved an average F1-Score of 0.8353. Our model also outperformed other models available in the literature.