Abstractive Summarizer using Bi-LSTM

Abstract

Abstractive Summarization (AS) of texts is the task of abstracting crucial information from the source. This paper presents an approach for text summarization in abstractive form with deep learning techniques. This paper develops a model that produces more precise and coherent summaries without redundancy problems. An efficient summarizer should provide the context from the input text in a brief manner. Thus, the output of the summarizer is abstracted information and is presented as a summary to the user. The dataset CNN Daily Mail is often used for multi -sentence summarizing techniques, and the AS models are usually used under an immense deep learning technique termed as seq-to-seq model. In the summarization part, the encoder-decoder model is typically applied. The most often used metric for evaluating the quality of summarization is identified: Recall - Oriented Understudy for Gisting Evaluation (ROUGE). The proposed summarizer performs better in terms of ROUGE.