Mining social media communities and content

摘要: Social Media is changing the way people find information, share knowledge and communicate with each other. The important factor contributing to growth of these technologies ability easily produce “user-generated content”. Blogs, Twitter, Wikipedia, Flickr YouTube are just a few examples Web 2.0 tools that drastically Internet landscape today. These platforms allow users annotate content more importantly, empower them information their social network. Friends can in turn, comment interact producer original also Such interactions foster communities online, media systems. User-generated graph thus two essential elements any system. Given vast amount user-generated being produced day easy access graph, how we analyze structure data understand nature online communication collaboration applications? This thesis presents systematic study through combined analysis its special properties, content. First, have developed framework for analyzing effectively. BlogVox opinion retrieval system large scale blog indexing engine. For given query term, retrieves ranks posts expressing sentiments (either positive or negative) towards terms. Further, index semantically syndicated1 feeds from news websites. We use sophisticated natural language processing system, OntoSem [163], stories build rich fact repository extracted real-time feeds. It enables other applications benefit such deep semantic by exporting text meaning representations Semantic language, OWL. Secondly, describe novel algorithms utilize properties graphs detect media. Communities an element systems detecting membership critical several real-world applications. Many community detection computationally expensive generally, do not well networks. In this work present approach benefits scale-free distribution node degrees extract efficiently. sites frequently provide additional meta-data about shared resources, usually form tags folksonomies . new algorithm combine structural obtained effectively communities. demonstrate availability properties. Finally, user perspective. first population subscribes organizes they read. has revealed interesting characteristics consume information. what now known as “feed distillation” task, which involves finding relevant term. Based on our understanding feed subscription patterns built prototype provides recommendations subscribe measures readership-based influence blogs different topics. We measure usage relatively phenomena called Microblogging. Microblogging current status short distributed instant messages, mobile phones, email Web. study, observations microblogging intentions studying content, topological geographical immediate talk daily activities seek The course research highlighted challenges presents. class problems requires us re-think mining, analysis. Comprehensive allows validate theories sciences psychology, but much larger than ever imagined. Ultimately leads better today future. 1RSS/ATOM