《User-Level Sentiment Analysis Incorporating Social Networks》.pdf

User-Level Sentiment Analysis Incorporating Social Networks Chenhao Tan Lillian Lee Jie Tang Dept. of Computer Science Dept. of Computer Science Dept. of Computer Science Cornell University Cornell University Tsinghua University chenhao@ llee@ jietang@ Long Jiang Ming Zhou Ping Li Microsoft Research Asia Microsoft Research Asia Dept. of Statistical Science Microsoft Corporation Microsoft Corporation Cornell University pkujianglong@ mingzhou@ pingli@ ABSTRACT termine viewpoint would enable users to make sense of the enor- We show that information about social relationships can be used to mous body of opinions expressed on the Internet, ranging from improve user-level sentiment analysis. The main motivation behind product reviews to political positions. our approach is that users that are somehow “connected” may be We propose to improve sentiment analysis by utilizing the infor- more likely to hold similar opinions; therefore, relationship infor- mation about user-user relationships made evident by online social mation can complement what we can extract about a user’s view- networks. We do so for two reasons. First, user-relationship infor- points from their utteran