Implementasi Metode Maximum Marginal Relevance (MMR) dan Algoritma Steiner Tree untuk Menentukan Storyline Dokumen Berita

Abstract

In chronological search process a news event online which using search engines, users must access the various sites that have relevance for the event. This is because search engines do not provide search results in a structured method. The search process requires conformity result of the query is entered as search criteria. Relevance textual news can be obtained by using a cosine-similarity summary of news by implementing the method of Maximum Marginal Relevance (MMR) which is determined based on the similarity to the query. In the same context, the search over 90 samples of news documents applied algorithm Steiner Tree in determining the best path in a collection of news documents (vertices) connected as a weighted similarity (edge) while the side directed toward a specific document (directed arc). Based on usability testing methods directed against six respondents as much as 66.7% of respondents considered that this application tasks more quickly and 75% of users were able to overcome the existing problems with the process of finding a time-saving news. From the performance testing of the algorithms applied, obtained by the complexity of the implementation of the MMR algorithm is O (n2) and Steiner Tree algorithm is O (n). Thus, the implementation of MMR and Steiner Tree in search applications of chronological events can reduce the level of absurdity news structurally. So as to facilitate the readers in determining the understanding of events happening quickly. Index Terms— text-summarization, MMR, steiner tree, Storyline, Artificial Intellegence (AI).