Time-aware Community-enhanced Social Information Retrieval

点击翻译按钮获取中文摘要

Search engines are the foremost means of information retrieval (IR) in the modern era, yet they have difficulty searching into knowledge repositories. This challenge arises not only due to the four V's of big data but also because they are not tailored to the users' differing information needs at different points in time. On the other hand, online social network (OSN) platforms are emerging as an indispensable means of communication, allowing for the seamless flow of information through user interaction (microscopic), which can result in viral spread effects (macroscopic). Social IR is to ease IR systems by integrating additional social network contexts for enhancing search engine rankings. Successfully as they are, Social IR methods, however, face three significant challenges which impede the optimum synergistic integration of OSN into IR: i) granularity: thus far, Social IR methods have been integrating social context at the fine-grain user level. As such, they are not scalable to a large number of users. Also, not all users actively contribute in OSNs; hence, they lack a rich social context that renders the user-level modelings unreliable and moot. Further, studies at the user level raise privacy concerns due to their intrusive approaches. Collective use of social information is also error-prone since it overlooks the locally dense communal inter-user associations, ii) temporality: user's information needs and her subjective definition of relevance in terms of topicality, novelty, reliability, understandability, importance, and scope are changing within time. While temporality has been well-explored in IR and social network analysis, the impacts of temporal OSN on IR processes has not explored yet, iii) balance: although propagating social information into the IR processes can enhance retrieval performance, Social IR may not always hold promise, e.g., when the search results solely rely on and overfit to the user's social model as opposed to her information needs. The overarching theme of my research is to study the possibility of developing temporal community-based Social IR techniques that would benefit from the OSNs for enhancing the search efficacy and efficiency. This proposal is original and novel because my students and I are the first, to the best of our knowledge, who fill the gap in Social IR by considering community structures and temporal dynamics in the OSNs as well as their synergistic trade-offs on IR processes via 1) scalable, 2) reliable, 3) non-intrusive, and 4) temporal methods, which would lead to a next-generation of IR technologies. The proposed research will provide a solid foundation for the training of HQP, including 3 Ph.D., 6 M.Sc., and 5 undergraduate students, who will acquire training in and become adept at social network analysis and information retrieval. The HQP can immediately be absorbed by the lively and growing data science and the analytics job market in Canada.