Literature Review of Automatic Text Summarization: Research Trend, Dataset and Method

Automatic text summarization can be defined as the process of presenting one or more text documents while maintaining the main information content using an automatic machine with no more than half the original text or less than the original text. Research in the field of text summarization began in the 1950s and until now there is no system that can produce summaries such as professionals or humans. This paper aims to identify and analyze methods, datasets and trends in automatic text summarization research from 2015 to 2019. The method used a systematic literature review (SLR) about automatic text summarization. The results obtained are that research on automatic text summarization is still relevant to date. The extractive approach is still in demand in the past three years because the extractive is easier than abstractive and the opportunity to combine methods is still open, for example using a neuro computing approach, namely the emergence of a new DQN method (Deep Q-Network) which shows comparable results and even better. The text summarization research trend has also undergone a slight change in the past three years where new things have emerged that are trends that are leading to optimization, how to optimize text summarization performance in order to get high accuracy.