The goal of this collaborative exploratory research project (0738652, Clement Yu, UIC and 0738727 Weiyi Meng, SUNY Binghamton) is to investigate the potential benefit of a dictionary-based approach for document retrieval. This project aims to demonstrate within one year that within a given domain, the use of multiple dictionaries (both domain specific dictionaries and general dictionaries) will obtain high retrieval effectiveness. The demonstration will be facilitated using existing TREC query and document collections. Rules for adding semantically related terms to queries that can yield high retrieval effectiveness within the selected domain will be identified. These rules will be carefully analyzed to obtain fundamental insights for the gain of effectiveness achieved due to the use of the dictionaries. This project is also to seek evidence that certain dictionary constructs (e.g., the frequency of use of a word in a context) can lead to significant gains in effectiveness if they are added to a dictionary. The research results from this project will lay the foundation in achieving longer-term goals that include the identification of domain independent principles of using different types of dictionaries in the same system, and the development of tools to add useful dictionary constructs across dictionaries and to assist users in query expansion semi-automatically when there are multiple dictionaries. It is expected that the proposed project can have a significant impact on search engine technology, including retrieval in specialized domains such as law and medicine, question-answering, blog retrieval and analysis, and enterprise search. Research results will be incorporated into several courses the PIs teach and students will be recruited to participate in the research activities. Research results will be disseminated through published papers as well as a textbook on Web-based search technology. The project Web site (www.cs.binghamton.edu/~meng/DocRetSGER.html) provides access to research results.
