CWU Imaging Lab
Research | Publications | Software | Links | People | Contacts | CWU | Internal |

Virtual Experts



 

Methodology and tools for creating Virtual Experts

BAA Number NMA401-02-BAA-0002

NIMA University Research Initiative (NURI)

Primary Research Area: Computer Science
(1) Knowledge Management/Virtual Experts

Secondary Research Area: Geospatial Information Science
(2) Data Representation, (3) Knowledge Development.

Principal Investigator
Dr. Boris Kovalerchuk


Dept. of Computer Science

ABSTRACT

This project addresses the knowledge management issues of NIMA's future operating environment including: more complex and detailed intelligence questions, need of remote access to multidisciplinary experts consultation, expert time and schedule constrains to deliver a reliable intelligence to policymakers and commanders.

Our primary research objectives are:
  1. Determine how to build knowledge-based and expert systems for use in supporting imagery analysis,

  2. Create tools to assist the knowledge engineer capture domain specific information and build the knowledge base,

  3. Incorporate the use of semantics into a knowledge management environment,

  4. Integrate multimedia information in a user-friendly, human-computer interface,

  5. Create tools to foster intelligent consultation of the virtual expert’ knowledge base.
Approach

The study in this project is based on the concept and tools of modern knowledge management ontologies, consistent data and expert mining. The major innovative components of this approach are:
  1. Extracting rules from data using relational data mining methods, which present a further development of methods known as inductive logic programming methods. These methods extract rules directly in contrast with neural networks that require an additional effort for converting the network to rules.

  2. "Extracting" rules from experts ("expert mining") using original interactive procedures with dynamic set of questions, where next questions depend on answers for previous questions. Also the set of questions is a minimal for the worst-case scenario.

  3. Testing contradiction between rules extracted from data using data mining techniques and "extracted" from experts using "expert mining" techniques.

  4. Deconflicting rules obtained from different sources by deleting parts that are in conflict and testing against trusted sources and cases using case-based reasoning,

  5. Capturing expert knowledge on the fly, when the experts work and record it using Image-DAML language.
We are building:
  1. Methodology for building virtual experts for use in supporting imagery analysis (specifically conflation) faster and more complete than with currently available technology.

  2. Tools to assist the knowledge engineer capture domain specific information and build the knowledge base.

  3. Tools for the synergistic use of multiple expert's knowledge so that the whole are greater than the part.

  4. Tools to integrate multimedia information in a user-friendly, human-computer interface.

  5. Experimental Knowledge Base for Imagery Conflation.


Research | Publications | Software | Links | People | Contacts | CWU | Internal |