Modul Word Graph Kata Benda pada KG_EDITOR Berbasis Java Desktop

Abstract

One natural language processing method that can represent result of semantic analysis of a text as a graph is Knowledge Graph (KG). BogorDelftConstruct is a software developed to analyze a text and visualize it as a graph using KG. BogorDelftConstruct was developed using MATLAB programming language, which is not freely accessible. Because of that, in this reseach we develop a system using Java programming language, so everyone can run this system freely. The objective of this research is to add a word graph module for Indonesian noun in the system. Indonesian nouns comprise of 20 word graph patterns. All but one of them are formed by affixation. The process of forming word graph is started by getting root word and affix contained in input word using stemming based on Kamus Besar Bahasa Indonesia (KBBI). The root word is used to determine the part-of-speech in KBBI, which is used as a parameter to determine which pattern is appropriate. The pattern containing the meaning of the word would then be generated by the system. In this word graph module, only 19 patterns can be implemented. From 176 words used in testing, there were only 4 errors, equivalent to an accuracy of more than 97%