ITR/IIS: Querying Heterogeneous and Uncertain Information

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

There are many different sources from which relevant information can be obtained for an intelligence task at hand, such as an analysis of operations planning by Al Qaeda. The accuracy of this information is frequently imperfect. Information obtained from each source is frequently also different in structure and scope. XML has emerged today as the standard for information exchange. Its flexibility permits the representation of information from heterogeneous sources in a single unified framework. The goal of this project is to manage uncertain (probabilistic) data, represented in XML. While there have been previous efforts to develop probabilistic relational systems, the need for representing and querying uncertain information is much greater for loosely structured data not readily amenable to a relational representation. XML data poses several modeling challenges that are the focus of this project: due to the data having (an inclusion hierarchy) structure, probabilities cannot be assigned independently; due to XML elements occurring at multiple granularities, probabilities may be assigned at multiple levels; due to the possibility of missing and repeated sub-elements, one may not be able to obtain complete distributions.