Co-Reference Resolution for the Indonesian Language Using Association Rules

Abstract In this paper, we proposed a co-reference resolution method for texts in the Indonesian language. The objective of co-reference resolution is to identify equivalence between entities as well as between pronouns and entities that were recognized in a named entity recognition phase. We propose a method that uses association rules. The method combines several features, such as pronoun and name classes, string similarity and position in the text, into a vector of attributes. Applied to a corpus of newspaper articles in the Indonesian language, the method yields an FMeasure of 84.12%. we compare the result to one of state-of-the-art machine learning method for co-reference resolution, decision tree, and the result is comparable.