Extractive Text Summarization Using Ontology and Graph-Based Method

Abstract

In recent years, many people started to take care of the physical health. The biomedical article is the trendy issue at the moment leading to the huge amount of knowledge created rapidly. In this paper, we propose a new automatic extractive text summarization technique based on graph representation that makes use of the Unified Medical Language System (UMLS), an ontology knowledge from the National Library of Medicine (NLM). We combined the graph building rules with a distance function between text documents, called Word Mover's Distance. To prioritize the core sentences, we extracted the summary by using a popular graph-based method from Google, PageRank. We compared our results with other text summarization software using 400 biological review papers as a corpus randomly sampled from PubMed Central. Our approach outperformed the baseline comparators in terms of Recall-Oriented Understudy for Gisting Evaluation (ROUGE) scores.