Deductive Composition of Multiple Data Sources

Principal Investigators:  Peter A Jarvis , Richard J Waldinger Mailing address: AI Center SRI International 333 Ravenswood Avenue Menlo Park, CA 94025-3493 AIC Program:  Representation and Reasoning

A framework is being developed for answering queries using multiple data sources and distributed software, based on techniques from deductive program synthesis. Answers need not be provided explicitly by any one data source, but may be inferred and computed from fragments provided by multiple data sources. The framework provides a new more flexible approach to interoperability.

The approach is being applied to composing data for geographical query answering and ecological modeling. In particular, deductive methods are being used to coordinate NASA satellite imagery from EOSDIS with SRI’s TerraVision system and with gazetteer sources. The effort depends upon a number of SRI technologies, including the following.

• SNARK: constitutes the deductive component of the proposed project, is a logically complete resolution-, paramodulation-, and rewriting-based theorem prover for first-order logic. It has a highly evolved sort structure, which allows us to represent taxonomic information concisely and to infer its consequences quickly. It has built-in facilities for fast temporal reasoning.

• Open Agent Architecture (OAA): makes it possible for data sources and software services distributed over the web to be invoked and coordinated through a single facilitator. SNARK can invoke procedures via OAA. EOSDIS, TerraVision, GeoWeb, web agents, and many of the sources and software we use in this project will be invoked via OAA.

• TerraVision: a distributed, interactive terrain visualization system. It allows users to navigate, in real time, through a 3-D graphical representation of a real landscape created from elevation data and aerial images of that landscape. TerraVision can browse huge datasets, which contain terabytes of data. All data can be distributed over multiple servers across the Web. 3-D VRML and GeoVRML models can be overlaid, e.g. buildings, wind vectors, etc. Datasets of different size, resolution, source, and coordinate system are supported.

• Artifact theory: an SRI effort to develop axiomatic theories of the physical world, so that physical devices and software will be able to communicate and cooperate. It includes a detailed theory of temporal, spatial and geographic reasoning. It is being developed in the language of SNARK, so that inferences based on axioms in the theory can be performed automatically.

As part of this project, we are building OAA agents to interface with various data sources and services that exist on the Web. These include the CIA Factbook, the UCSB Alexandria Digital Library Gazetteer, and OpenGIS Consortium (OGC) Web Map Servers, among others.

	Open Agent Architecture The Open Agent ArchitectureTM (OAA®) is a framework for integrating a community of heterogeneous software agents in a distributed environment.
	TerraVision TerraVision is a distributed, interactive terrain visualization system developed by SRI International. It allows users to navigate, in real time, through a 3-D graphical representation of a real landscape created from elevation data and aerial images of that landscape. All data can be distributed over the Web and support is provided for overlaying GeoVRML 3D models.

Showing most recent 5 out of 6  [View All]

• Waldinger, R. and Appelt, D. E. and Fry, J. and Israel, D. J. and Jarvis, P. and Martin, D. and Riehemann, S. and Stickel, M. E. and Tyson, M. and Hobbs, J. and Dungan, J. L. Deductive Question Answering from Multiple Resources. in New Directions in Question Answering, AAAI, 2004.  [PDF, Details]

• Waldinger, R., Jarvis, P., and Dungan, J. Program Synthesis for Multi-Agent Question Answering. in International Symposium on Verification (Theory and Practice); Festschrift celebrating Zohar Manna’s 64th Birthday, Springer Verlag, Lecture Notes in Computer Science, July 2003.  [PDF, Details]

• Waldinger, R., Jarvis, P., and Dungan, J. Pointing to Places in a Deductive Geospatial Theory, in Workshop on Analysis of Geographical References; Human Language Technology Conference , Edmonton, Canada, pp. 10–17, May–Jun 2003.  [PDF, Details]

• Waldinger, R. , Jarvis, P., and Dungan, J. Using Deduction to Choreograph Multiple Data Sources, in Semantic Web Technologies for Searching and Retrieving, Sanibel Island, Florida, Oct 2003.  [PDF, Details]

• Waldinger, R., Reddy, M., Culy, C., Hobbs, J., and Dungan, J. Deductive Response to Geographic Queries, in GIScience 2002, Boulder, CO, Sep 2002.  [PDF, Details]