SBIR-STTR Award

We propose to develop an intelligent research assistant system for harvesting and display of knowledge from the World Wide Web. The proposed application, IRIA (Internet Research Intelligent Assistant), aids and automates a user's search for information by autonomously querying multiple knowledge resources, by organizing and presenting information as it is found, by modifying and extending the ongoing search based on unobtrusive inspection of user browsing, and by integrating data from research sessions into a knowledge base of relevant information resources. IRIA employs a human-like model of knowledge-base search, traversing a semantic map of available information resources in an autonomous and context-sensitive way. Innovative machine learning techniques are then applied to collate and analyze retrieved information for presentation to the user. The combination of these two techniques allows both "fire and forget" web search, in which the system collects and analyzes information offline, and "incremental" web search, in which user browsing of retrieved pages is used to automatically adjust the search in a context-sensitive way. The results of these searches and user responses are stored in the semantic map, both to improve the results of future queries and to provide a searchable directory to the information resources available

The objective of the proposed project is to develop an artificial intelligence toolkit that provides intelligent information management capabilities for interactive knowledge-intensive applications. The explosion of information in the modern environment demands the ability to collect, organize, manage, and search large amounts of information across a wide variety of real-world applications. Traditional large-scale database and keyword-based search systems are inadequate at providing ready access to relevant knowledge. A new approach is needed to information management that provides a user or a workgroup with the ability to find, add, and use information quickly based on a personalized "map" of information relevant to them. A system based on a model of human memory can organize information into a "knowledge map", which enables "reminding engines" which proactively find and recommend information useful to a working user and "workgroup memories" that learn from user behavior to improve the knowledge map over time. The toolkit will enable developers to embed these facilities in a wide variety of applications, including workgroup, internet, intranet, and education and training systems. Available pervasively in a user's work environment, knowledge maps will empower users to get to information they need quickly and effortlessly.

Keywords:
ARTIFICIAL INTELLIGENCEINFORMATION MANAGEMENT KNOWLEDGE MANAGEMENT CLIENT-SERVER APPLICATIONS ADAPTI