Ekstraksi Informasi Semantik dan Spatiotemporal pada Artikel Online Terkait Bencana di Indonesia

Abstract

Indonesia is one country that has a high risk of natural disasters. Ranked in 36 out of 172 countries in disaster index and having 2,372 disaster incidents in 2017, actions need to be taken to minimize the impact of natural disasters. One of it is to do a hazard map modeling. In making hazard maps, several approaches can be used, one of which is the semantic approach to extract disaster information. Therefore, this study aims to develop a system that can be used to extract spatiotemporal and semantic information related to natural disasters in Indonesia. This study uses the NLP method in conducting the information extraction process and  carried out using the GATE (General Architecture for Text Engineering) application. In processing Indonesian language articles, it is necessary to develop the plugin because the Indonesian information structure is different from the default information structure in GATE application. The plugin development process is done by using ontology as the basis for determining semantic information. Literature study was carried out related to government regulations that further explained the need for semantic and spatiotemporal information about disaster events. system performance developed produces a precision value of 38% and a recall value of 32%. this is because the system experiences some difficulties in carrying out the information inference process. The reason for low precision rate is because the rules used in the inference process to pair the three types of information still cannot accommodate the variation of information positions in different sentences.