As the Navy's vision of FORCEnet for network centric warfare is realized, the analysts and warfighters alike will have access to a multitude of structured and semi-structured information sources. To ensure that this wealth of information does not turn into an information overload, human decision analysts will require information mediators to increase the efficiency of their access to the information sources. To be effective in such decision support scenarios, the mediator frameworks need to be highly adaptive to the information sources they mediate over, as well as the users they serve. Unfortunately, most existing mediators operate with very little knowledge about the organization and contents of the sources, as well as the higher level goals of the user. As a result, their query processing is often inflexible. In this research, we propose to address critical issues in adaptive information integration. Our proposed research will make three key contributions to information integration for decision support: Adaptive Source Selection & Query Planning: Designing mediators that are adaptive to the source profiles as well as user needs in query planning. In particular, mediators that can flexibly support multiple and often conflicting objectives (e.g. between latency, quality and completeness) that the users have in selecting information sources (query plans). Adaptive Ranking & Retrieval: Designing mediators that can handle query imprecision and data incompleteness (and inconsistency). This is needed because, most data sources tend to have incomplete records (e.g. due to missing attribute values), and most lay users combine querying and browsing and often pose imprecise queries. Adaptive User Interaction: Designing mediators that are sensitive to the lay users and pro-actively take usability considerations into account. This will include explaining the results to the users, and helping the users in formulating their queries. These proposed tasks are all closely related to the "information integration optimization" thrust (126.96.36.199) of the BAA (and also have connections to the ''uncertainty management and data refinement" thrust (188.8.131.52)). The proposed research aims to combine insights and techniques from several traditional disciplines-including databases, information retrieval, machine learning, preference elicitation and automated planning and plan recognition to address foundational issues of adaptive information integration. The proposal is backed by the PI's strong track record in AI and Database research and education, as well as his successful track record in mentoring highly qualified graduate students in information integration. We believe that the adaptive information integration framework developed in this research will be particularly relevant in the information gathering scenarios faced by military and intelligence analysts. We intend to demonstrate this in the context of the Navy information integration needs by adapting our prototype to the type of data sources found in command/control and intelligence analysis scenarios-including relational, text-oriented and semi-structured sources.
|Effective start/end date||10/1/08 → 9/29/12|
- DOD-NAVY: Office of Naval Research (ONR): $472,587.00