Peringkasan Multi-Dokumen Berita Bahasa Indonesia Menggunakan Algoritma Lexrank

Abstract

The rapid growth of technology makes its users increase. The many conveniences offered by a technology can ease human work. One of them is the ease of accessing information via the internet. The habit of Indonesian people in reading print media has shifted to digital media. News is information about the occurrence of a hot event that is presented through the mass media. The proliferation of digital news portal developments is not comparable to the provision of quality news content. Some news have the same information but are written with different titles and narratives. The style of news which is convoluted in words and sentences makes it difficult for readers to find the essence of a news story, so that readers need a lot of time to read the news. Therefore, an approach is needed to summarize multi-text Indonesian news documents to get the gist of news stories that have the same subject without having to read the entire news text. This multi-news document summary use the lexrank algorithm, which use the TF-IDF weighting system to find important words in each sentence and then calculates the vector length of each sentence. The lexrank algorithm uses cosine similiarity to compare one sentence to another. After that, the sentences will be ranked based on the highest value to be used as a summary. Next, the results of the system summary will be compared with the expert summary using ROUGE. This study uses 33 raw news data to be tested using the lexrank algorithm. Testing this method only uses 3 news from each news portal (detik.com, cnbcindonesia and antaranews) that has been determined so as to produce recall, precision and F-score of 0.693, 0.6 and 0.682, respectively.