Computer Science Building, Room 151
Social applications on the Web have appeared as communication spaces for sharing knowledge and information. In particular, social applications can be considered valuable information sources because information in the applications is not only easily accessible but also revealing in that the information accrues via interactions between people.
In this work, we address methods for finding relevant information in social media applications that use unique properties of these applications. In particular, we focus on three unique structures in social media: hierarchical structure, conversational structure, and social structure. Hierarchical structures are used to organize information according to certain rules. Conversational structures are formed by interactions within communities such as replies. Social structures represent social relationships among community members. These structures are designed to organize information and encourage people to participate in discussions in social applications. Accordingly, contexts extracted from these structures can be used to improve the effectiveness of search in social media relative to representations based solely on text content.
To exploit these structures in retrieval frameworks, we need to address three challenges as follows. First, we should discover each structure because it is often obscure. Second, we need to extract relevant contexts from each structure because not all the contexts in a structure are relevant for retrieval. Last, we should represent each context or their combinations in a representation framework so that they can be encoded as retrieval components such as documents. In this work, we introduce an effective representation framework for multiple contexts. We then discuss how to discover or define each structure and how to extract relevant contexts from the structure. Using the representation framework, these relevant contexts are integrated into retrieval algorithms. To demonstrate that these structures can improve search in social media, the retrieval models and frameworks incorporating these structures are evaluated through experiments using data collections gathered from a variety of social media applications.
In addition, we address two minor challenges related to social media search. First, it is not always easy to find relevant information from relevant objects if the objects are large. Accordingly, we address identification of relevant substructures in such objects. Second, text reuse structures are important since these structures have the potential to affect various retrieval tasks. In this thesis, we introduce text reuse structures and analyze text reuse patterns in real social applications.
Advisor: W. Bruce Croft