Distributed Relevance Ranking in Heterogeneous Document Collections

72223-Research professionals need advanced search and retrieval technology to identify, assemble, and rank-for-relevance scientific and technical information resident on the Internet. This requirement is particularly challenging when utilizing the deep web in which most of the important data and documents reside. Currently available software tools are inadequate. This project will develop the key components of an automated distributed approach to ranking and presenting the results of intensive searches for scientific and technical information identified from multiple source collections housed in the deep web. Phase I demonstrated the feasibility of developing state-of-the-art document ranking algorithms by extensive testing against manual ranking by human experts, in accordance with accepted criteria. Three powerful ranking algorithms that operate at increasing levels of sophistication were successfully developed. In Phase II, the capabilities of these ranking algorithms will be embodied into a sophisticated software package. In response to a user¿s query, the package will identify, retrieve, and rank the most relevant documents across a large number of distributed collections of heterogeneous documents that are resident in the deep web. Commercial Applications and Other Benefits as described by awardee: The software package should be deployable in both grid and non-grid computing environments. More efficient, faster, high-quality, cost-competitive search and retrieval of highly valuable information resident in the deep web would provide information-centric government agencies and commercial enterprises with measurable performance advantages.