Dependency Syntactic Analysis for Opinion Mining

Background/Objectives: Dependency parsing is one of the remaining problems that attracts researchers’ interest because of its contribution to natural language processing applications. This paper presents a dependency parser and provides an effective sentiment analysis model for Vietnamese. Methods/Statistical Analysis: Up to now, there has been some remarkable work on dependency parsing for Vietnamese, of which the approach of transforming the constituents into a dependency grammar has been proven to give the best performance. In this study, we proposed another way to parse Vietnamese sentences to create dependency trees based on the incremental algorithm by Covington and Graph Unification Programming. We have also combined a dependency parser, sentiment dictionary, domain ontology, and syntactic dependency rules for solving the sentiment analysis problem. Findings: We found that the experiments using simple sentences collected from Vietnamese reviews provided better results than with other methods. The combination of dependency parser, sentiment dictionary, domain ontology, and syntactic dependency rules may be helpful when dealing with a sentiment analysis problem. This model can be applied to Vietnamese and other languages. Application/Improvements: The positive results, which came from an experimental period, will provide us with a basis from which to expand grammar rules to build a parser for more complicated Vietnamese sentences and to integrate it into many cognitive tasks, such as sentiment analysis.