Pair Extraction of Aspect and Implicit Opinion Word based on its Co-occurrence in Corpus of Bahasa Indonesia

Abstract

In implicit opinion sentences containing implicit opinions, the use of traditional dictionary-based method is no longer effective, since there is no any opinion word serving as a clue to identify the sentiment value. This study proposes an approach to obtain pairs of aspect and opinion word from sentences containing implicit opinion. Two methods are developed this approach. The first method serves to separate clauses from a compound implicit opinion sentence and refine its corresponding parse-tree from clauses as a result of the separation. The second method determines the opinion word from the implicit opinion clause. This method uses co-occurrence of aspect and opinion word based on a corpus of explicit opinion sentences. In this study, an initial experiment was conducted using a data set containing 30 implicit opinion sentences with 76 clauses. The proposed approach is capable of detecting pairs of aspect and opinion from the given corpus of Bahasa Indonesia.