BUILDING A RESTAURANT ASSESSMENT SYSTEM IN THUA THIEN HUE PROVINCE BASED ON ONLINE COMMENTS

Abstract

Vietnamese opinion mining systems are based on the lexicon-based approach using the VietSentiWordNet dictionary. However, this data dictionary applies to the news domain, so when used to classify in the tourism domain, it will be ineffective and easy to cause confusion. The objective of this paper is to build a restaurant assessment system with high classification efficiency in the tourism domain. To build the system, we use lexicon-based approach to opinion mining combined with the Vietnamese opinion dictionary in the tourism domain VietSentiWordNetPlus. In addition, we also apply data preprocessing techniques to the comments to increase the semantics of the sentences. The experimental results showed that, our system gave better opinion classification results, with average accuracy, precision, recall and F-score 84.64%; 76.39%; 81.12%; 78.15% versus 71.76%; 63.64%; 68.72%; 63.82% of the system uses the VietSentiWordNet dictionary. Our system is highly effective when classifying opinion with data sources in the tourism domain such as restaurants, hotels, tourist attractions.