Detecting New Connections via Social Media Content Analysis

Recently, as a part of continuous efforts of exploring social media to support serendipity, we considered the possibility of finding potentially interesting new connections via social media content analysis. In particular, we are interested in spotting non-existent connections among Twitter users who share certain interests in their twitter posts. We assume the rich contents of the social media can provide good opportunities for finding interesting and surprising connections.

It is well known that the existing users’ profile from social media can be used to match people and make recommendations. In addition, user interactions and situational information available via social media tools, such as location, proximity, time etc, have also been used to analyse people connections. We wonder whether or not the contents of social media posts, i.e. what people actually talk about over time, can provide additional information for finding useful links.

In fact, content analysis of social media has recently been receiving an increasing attention. For example, Carter et al. (2004) connect loosely coupled research groups by analysing and matching their interests based on their emails. Matsumura et al. (2005) mine social network from a message board based on the terms propagating via messages among the users. Canini et al.’s (2011) study on Twitter topical relevancy shows that “the credibility of a Twitter account with respect to a particular domain depends in large part on the strength of association between the textual content of the account and the domain in question”. Chelmis et al. (2011) also suggest that, for social networking analysis, the social networking content analysis is important in addition to the graph analysis.

With regard to the types of user connections, we are particularly interested in potential connections of Twitter users who are not already in contact, either in the form of following or interaction. Although interesting connections may well occur between friends and acquaintances, we believe it would more likely to bring interesting and surprising connections between strangers, particularly based on what they are talking about on Twitter.

Another issue we investigate is the relevance and usefulness of different sources of information for the connection detection. In our study, two main sources of information are considered: the tweets and the web pages hyperlinked from the tweets. Our earlier experiment showed that these different sources of information may have different contribution to the detection of serendipitous interests of twitter users, and we suspect that these sources may allow us to find different types of connections among the users.

We are testing to detect the Twitter user connections employing NLP content analysis techniques and statistical similarity measuring algorithms, and our initial experiment showed some supportive evidence for this approach. For example, we could detect some new connections which show closer semantic relevance than existing ones. We assume that recommendation of such new connections can potentially lead to serendipitous encounters.

References:

Canini, Kevin R, Bongwon Suh and Peter L. Pirolli (2011). Finding Credible Information Sources in Social Network Based on Content and Social Structure. In Proceedings of the 2011 IEEE International Conference on Social Computing (SocialCom 2011), Boston, US.

Carter, Scott, Jennifer Mankoff, P. Goddi (2004). Building Connections among Loosely Coupled Groups: Hebb’s Rule at Work. Computer Supported Cooperative Work 13(3), pp. 305-327.

Chelmis, Charalampos and Viktor K. Prasanna (2011) Social Networking Analysis: A State of the Art and the Effect of Semantics. In Proceedings of the 2011 IEEE International Conference on Social Computing (SocialCom 2011), Boston, US.

Matsumura, Naohiro, David E. Goldberg and Xavier Llorà (2005). Mining directed social network from message board. In Proceedings of The 14th International World Wide Web Conference (WWW2005), Chiba, Japan.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

 

In other news