Reinforcement Syntactic Dependency Tree Reasoning for Target-Oriented Opinion Word Extraction

Abstract

Target-oriented Opinion Word Extraction (TOWE) is a subtask of Aspect Based Sentiment Analysis (ABSA), which aims to extract fine-grained opinion terms for a given aspect term from a sentence. In TOWE task, syntactic dependency tree is useful as it provides explanation to identify opinion terms to the given aspect term. It is necessary to mine relationships between aspect and opinion terms for a better performance. Previous works introduced syntactic dependency tree into TOWE task but lacked of explicit explanation. In this paper, we propose a novel model named MM-TOWE, which leverages Monte-Carlo tree search to enhance Markov decision process (MDP) model for Target-oriented Opinion Word Extraction task. We formulate TOWE task as an MDP of reasoning over the syntactic dependency tree. By learning the dependency relationships between aspect terms and opinion terms, our model can reason a path for an explicit explanation. Extensive experimental results illustrate that our proposed model outperforms the state-of-the-art methods.