Chapter

Web-Age Information Management

Volume 6184 of the series Lecture Notes in Computer Science pp 471-482

Exploring the Sentiment Strength of User Reviews

  • Yao LuAffiliated withDept. of Computer Sci. and Tech., University of Sci. and Tech. of ChinaDepartment of Computer Science, City University of Hong Kong, HKSARJoint Research Lab of Excellence, CityU-USTC Advanced Research Institute
  • , Xiangfei KongAffiliated withDepartment of Computer Science, City University of Hong Kong, HKSAR
  • , Xiaojun QuanAffiliated withDepartment of Computer Science, City University of Hong Kong, HKSAR
  • , Wenyin LiuAffiliated withDepartment of Computer Science, City University of Hong Kong, HKSARJoint Research Lab of Excellence, CityU-USTC Advanced Research Institute
  • , Yinlong XuAffiliated withDept. of Computer Sci. and Tech., University of Sci. and Tech. of ChinaJoint Research Lab of Excellence, CityU-USTC Advanced Research Institute

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Existing research efforts in sentiment analysis of online user reviews mainly focus on extracting features (such as quality and price) of products/services and classifying users’ sentiments into semantic orientations (such as positive, negative or neutral). However, few of them take the strength of user sentiments into consideration, which is particularly important in measuring the overall quality of products/services. Intuitively, different reviews for the same feature should have quite different sentiment strength, even though they may express the same polarity of sentiment. This paper presents an approach to estimating the sentiment strength of user reviews according to the strength of adverbs and adjectives expressed by users in their opinion phrases. Experimental result on a hotel review dataset in Chinese shows that the proposed approach is effective in the task of sentiment classification and achieves a good performance on a multi-scale evaluation.

Keywords

opinion mining sentiment analysis sentiment strength text mining