Analyzing sentiments in Web 2.0 social media data in Chinese: experiments on business and marketing related Chinese Web forums

  • Authors:

  • Li Fan;Yulei Zhang;Yan Dang;Hsinchun Chen

  • Affiliations:

  • Department of Management Information Systems, Eller College of Management, University of Arizona, Tucson, USA 85721;Computer Information Systems, The W. A. Franke College of Business, Northern Arizona University, Flagstaff, USA 86011;Computer Information Systems, The W. A. Franke College of Business, Northern Arizona University, Flagstaff, USA 86011;Department of Management Information Systems, Eller College of Management, University of Arizona, Tucson, USA 85721

  • Venue:
  • Information Technology and Management
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

Web 2.0 has brought a huge amount of user-generated, social media data that contains rich information about people's opinions and ideas towards various products, services, and ongoing social and political events. Nowadays, many companies start to look into and try to leverage this new type of data to understand their customers in order to make better business strategies and services. As a nation with rapid economic growth in recently years, China has become visible and started to play an important role in the global business and economy. Also, with the large number of Chinese Internet users, a considerable amount of options about Chinese business and market have been expressed in social media sites. Thus, it will be of interest to explore and understand those user-generated contents in Chinese. In this study, we develop an integrated framework to analyze user sentiments from Chinese social media sites by leveraging sentiment analysis techniques. Based on the framework, we conduct experiments on two popular Chinese Web forums, both related to business and marketing. By utilizing Elastic Net together with a rich body of feature representations, we achieve the highest F-measures of 84.4 and 86.7聽% for the two data sets, respectively. We also demonstrate the interpretability of Elastic Net by discussing the top-ranked features with positive or negative sentiments.