Automatic Text Summarization (Ats) pada Berita Bahasa Indonesia Berdasarkan 5w1h dengan Metode Peringkasan Ekstraktif dan Maximal Marginal Relevance (Mmr)

Abstract

One of effective way to present a long text in a short form is extraction. Extraction is a process that aims to transform text into a formatted and concise structure but still retains the important points of the text. Long news texts make it difficult for readers to conclude important information from the news text. Therefore, an automatic summary system is needed so that readers can quickly get a summary and gist of the news text. To get this summary, methods such as Maximal Marginal Relevance (MMR) and Named Entity Recognition (NER) are needed. In this study Maximal Marginal Relevance (MMR) is used to remove redundancies from the initial news text whose summary results will be extracted into entities related to person names, times and places using Named Entity Recognition (NER). The data tested was in the form of news text from online news portals of 100 data. The results of this test produce an accuracy of 81% for the who entity, 83% for the where entity, and 93% for the when entity.