Graph-based Interactive Bibliographic Information Retrieval Systems

In the big data era, we have witnessed the explosion of scholarly literature. This explosion has imposed challenges to the retrieval of bibliographic information. Retrieval of intended bibliographic information has become challenging due to the overwhelming search results returned by bibliographic information retrieval systems for given input queries. At the same time, users' bibliographic information needs have become more specific such that only information that best matches their needs is seen as relevant. Current bibliographic information retrieval systems such as Web of Science, Scopus, and Google Scholar have become an unalienable component in searching bibliographic data. However, these systems have limited support of complex bibliographic queries. For example, a query- "papers on information retrieval, which were cited by John's papers that had been presented in SIGIR" is an ordinary information need that researchers may have, but is not appropriately representable in these systems. In addition, these systems only support search for papers and do not support other bibliographic entities such as authors and terms as the final search results. Therefore, in this dissertation, we propose several bibliographic information retrieval systems that can address complex bibliographic queries. We propose form-, natural language-, and visual graph-based systems that allow users to formulate bibliographic queries in a variety of ways. The form-based system allows users to formulate queries by selecting forms and input values in those selected forms. In the natural language-based system, users formulate queries using a natural language. Users formulate queries by drawing nodes and links in the visual graph-based system. These systems are based on a graph model to enhance retrieval efficiency and provides interfaces for users to formulate queries interactively. Through a system-centered evaluation, we find that our graph-based system took less time to process complex queries than a relational-entity-based system (two secs vs. several mins on average). In addition, our visual graph-based system can deal with the representation of advanced queries such as bibliographic coupling, paper co-citation, and author co-citation, while current bibliographic information systems do not support these queries. A user-centered evaluation reveals that participants rated the natural language-based system the most useful, easy to use, and easy to learn. Participants also reported that the form-based system was easier to learn than the visual graph-based system. Based on the results of a usability evaluation, we find that the form-based system is preferred for low-complexity tasks while the visual graph-based system is preferred for high-complexity tasks. The strength of the natural language-based system is that no additional effort is needed to formulate more complex queries. The proposed systems are effective and efficient solutions for addressing complex bibliographic information needs. In addition, we believe the experimental design and results shown in this paper can serve as a useful guideline and benchmark for future studies.