Abstract:
|
Data in online social network and social media systems provides a significant source of information about individual attitudes, preferences, and relationships. While there is a large body of work using statistical and machine learning techniques to analyze topics in document collections and propagation of information, these efforts typically do not try to jointly exploit the textual content and the relationships among users. In this work, we analyze the relationship structure, flow of discussion, and word usage in a large Twitter collection collected based on popular political hashtags. We investigate several machine learning methods for learning network and temporal patterns from these data, present an empirical evaluation of the algorithms, and discuss the implications of the discovered results for understanding behavior in Twitter. This work is supported by NSF grant DMS #1246818.
|