User Modeling and Personalization in the Microblogging Sphere

摘要: Microblogging has become a popular mechanism for people to publish, share, and propagate information on the Web. The massive amount of digital traces that have left in microblogging sphere, creates new possibilities poses challenges user modeling personalization. How can activities be exploited infer individual users' interests? semantically meaningful profiles constructed support different applications? Does behavior vary between cultural groups? What is impact strategies characteristics performance personalized recommendations? In this thesis, we answer research questions above introduce generic framework provides variety inferring interests from streams. We propose evaluate techniques allow exploiting external resources enrich semantics short messages. explore approaches deducing topics based enriched data. Furthermore, investigate various weighting schemes constructing incorporate temporal constraints into process. With flexible design choices, allows which consumed applications. apply our analyze across groups platforms. By characteristics, unveil key differences Chinese American users. Finally, context recommender systems. results analyses show are significantly influenced by alternatives. set experiments reveal semantic enrichment microposts consideration patterns improve also prove incorporation public trends domain specific knowledge process improves quality recommendations.