Who’s the campus stud: predicting sentimentleaders on a campus chat network
Muktabh Mayank SrivastavaMarch 17, 2012
In this paper, I have worked on the problem of tracing inﬂuentialpeople on the campus chat network. The campus chat has been modelledin form of a network using Natural Language Processing techniques andthen centrality algorithms of degree and betweenness have been appliedon the given network to get the sentiment leaders out.
Chat rooms and twitter like social networks have become a very important wayto express sentiment in the recent years. Thousand or millions of people canproject their views on these social media. Sentiment Leaders are people whoinitiate or control sentiments of the overall presence of people. Problem of ﬁnd-ing Sentiment leaders is very important due to many reasons. While consumerbased companies want to get positive reviews from sentiment leaders, tracingsentiment leaders also helps in tracing out terrorists and miscreants. Productsbased on such analysis like career-iq and IBM’s cellphone network analysis plat-form have seen billions of investments in last few years.The problem addressed ion this paper is ﬁnding out the sentiment leaders on thecampus chat network using Natural Language Processing (NLP) and NetworkAnalysis techniques. The campus chat log of 6 consecutive academic months(September 2010-December 2010) and February 2011 - March 2011 was mod-elled as a chat graph. This chat graph was then analysed for various kinds of centralities for ﬁnding out the leaders.
2 Making the Chat Network
Chat Network is made by analysing the chat content of ﬁle sharing network of campus and making a graph of the form a followed by b (to follow means to getinvolved in discussion initiated by). When a user types a line on the chatroom,he/she can either be desiring one of the three eﬀects:1