The increasing popularity of social networking websites has greatly facilitated expression of people’s personal opinions. However, in the current literature, opinions expressed in online communities are not taken into consideration when constructing and analyzing social networks. This paper is aimed at enhancing traditional social network analysis by incorporating opinions mined from online user-generated content. We present a new approach based on opinion-based social networks and a related PageRank-like method, named OpinionRank, to identify opinion leaders in social networks. This study combines social network analysis and opinion mining technology to study opinion networks and is aimed at answering the following research questions: a) How can one extract people’s opinions about each other and represent them in social networks? b) Can sentiment clues in opinions help to identify leaders of online communities? The main contributions of this work are as follows a) We introduced the concept of opinion networks and proposed a new approach for construction of opinion networks based on the implicit relations between social members. b) By analyzing the characteristics of opinion networks, this study proposed node-based and link-based ranking models to identify core members in such social networks. The effectiveness of different network sorting methods are evaluated on real word datasets. c) This study also found that the overall opinion orientation of the entire opinion network are positive, and most social members express more positive opinions than negatives. Large scale opinion networks are rich of sentiment information, and the complexity of large scale opinion networks makes them more stable than small networks.
修改评论