Automated text summarization of scientific documents

Abstract

Text summarization plays a major role in natural language processing, especially in scientific communities like researchers, students, and so on. Due to the number of scientific publications available online rapidly rising, it takes too much time to identify the most appropriate, quality, and relevant materials for their search out of thousands. Therefore, there should be an alternative way to sort out and simplify the search and get a quality and appropriate document based on our search. The aim of this work is to generate an online platform for a digital library that provides a good summary of any scientific document which is subscribed to by the library of the institution. Therefore, we need to find an appropriate and best suitable text summarization algorithm out of some state-of-the-art text processing algorithms such as the Text Rank algorithm, TF-IDF algorithm, and K-Means algorithm, which have been used in different text processing scenarios. To evaluate and select the best suitable algorithm, we used a publicly available scientific dataset and manually generated a summary from the dataset. From the experiments processed, the Text Rank algorithm performed better than the other algorithms.