Survey on Abstractive Text Summarization

This paper provides a review on some of the significant research work done on abstractive text summarization. The process of generating the summary from one or more text corpus, by keeping the key points in the corpus is called text summarization. The most prominent technique in text summarization is an abstractive and extractive method. The extractive summarization is purely based on the algorithm and it just copies the most relevant sentence/words from the input text corpus and creating the summary. An abstractive method generates new sentences/words that may/may not be in the input corpus. This paper focuses on the abstractive text summarization. This paper explains the overview of the various processes in abstractive text summarization. It includes data processing, word embedding, basic model architecture, training, and validation process and the paper narrates the current research in this field. It includes different types of architectures, attention mechanism, supervised and reinforcement learning, the pros and cons of different architecture. Systematic comparison of different text summarization models will provide the future direction of text summarization.