Automatic Extractive Text Summarization for Indonesian News Articles Using Maximal Marginal Relevance and Non-Negative Matrix Factorization

Abstract

The rapid development of the internet has led to the creation of many online news. Competition between the publishers usually will lead to the making of interesting and often misleading titles that are not conveyed the content of the news itself. To overcome this problem, we have developed a text summarization that extracts the keyword from the content, then used that keyword to produce a summary. This research combines Maximal Marginal Relevance (MMR) and Nonnegative Matrix Factorization (NMF) to build automatic extractive text summarized. MMR used to summarize the text automatically and NMF is used to extract the keyword to form the query for MMR. Based on our experiment, the performance of the summarizer system that combines NMF as the keyword generator produce better performance in ROUGE-1 and ROUGE-2 compared to systems that only use a title with MMR. We also perform an experiment to see the impact on the number of a keyword in the summary result. We also found that 8 keywords are the ideal number of keywords to represents the idea of the main text.