Intent of Work

Program: DARPA Agent Markup Language

Project: UML-Based Ontology Toolset (UBOT)

Contract No. F30602-00-C-0188

Period covered: February 2002 – September 2002

Prepared by: Paul Kogut

Lockheed Martin Management and Data Systems

[email protected]

610-354-3524

Date: January 31, 2001

Ontology Engineering Tools

Technical Goals:

Extend and refine the ontology engineering tools so that they can be used in real-world large-scale DAML applications. Major extensions are anticipated to support modeling of DAML rules and consistency checking of merged ontologies. Specific tasks include:

· Continue to refine the UML profile for DAML via collaboration with the GRCI DAML team and Telelogic. Continue to facilitate bi-directional technology transition between the DAML program and the Object Management Group.

· Investigate the use of alternative UML diagrams (e.g., activity diagrams) for modeling DAML-S

· Refine the semantic consistency checking capabilities of ConsVISor and add a debugging advice generation capability based on an ontology of DAML errors.

· Continue to provide candidate revisions to the DAML axiomatic semantics specification as a byproduct of formally modeling DAML with Specware and theorem proving with SNARK.

· Develop a consistency reasoning agent that integrates the Specware/SNARK and ConsVISor approaches.

· Extend and refine UML to DAML and DAML to UML translation

· Check ontologies for redundant concepts with WordNet synonym data

Metrics:

• % of known errors found (use test ontologies with manually seeded errors)
• % of errors found in ontologies that are legitimate errors
• processing time required to check ontologies related to size/complexity of ontology

DAML Experiment:

The UBOT team plans to apply ConsVISor and Specware/SNARK tools to check selected ontologies that are developed and merged by other DAML teams for the RDO/ONA domain. Errors will be reported to ontology developers and the lessons learned will be used to refine the consistency checking tools. The tools for UML-based visualization and generation of DAML ontologies will fit into the DAML Experiment as described below in the section on trial applications.

Next Logical Steps:

• Investigate tool support for ontology design patterns, ontology refactoring and other software engineering inspired innovations.
• Update tools to support UML 2.0.
• Integrate UBOT ontology engineering tools with the agent kit being developed by Lockheed Martin Advanced Technology Labs for the DARPA CoABS program.

DAML Annotation Tools

Technical Goals:

Extend and refine the DAML annotation tools so that they can be used in real-world large-scale DAML applications. The focus will be on evolving AeroDAML into a DAML-S annotated service for use by other DAML tools and applications. Specific tasks include:

• Design and develop an improved AeroDAML that allows the user to choose an ontology as the basis for annotation generation. Candidate ontologies include: OpenCyc, Cyc Upper Ontology, CIA World Fact Book, IEEE Standard Upper Ontology, Dublin Core, and UNSPSC.
• Design and develop a tool that allows the user to customize AeroDAML annotation generation by mapping the default ontology to a user-defined ontology.
• Develop and update a DAML-S description of AeroDAML
• Collaborate with the GRCI DAML team to integrate AeroDAML into their Ontology Driven Knowledge Dissemination (ODKD) application framework and integrate the GRCI ontology articulation service with the AeroDAML service.
• Collaborate with the Teknowledge DAML team to integrate AeroDAML into their MS Word-based annotation tool.
• Collaborate with the UMBC DAML team to apply AeroDAML to their IT Talks application (http://ittalks.org).
• Expand the annotation generation coverage of AeroDAML (i.e., extract more classes and properties) by integrating the AeroText knowledge base developed for the Horus Automated Markup Tool.
• Collaborate with the Horus team to apply the AeroText-based Automated Markup Tool to the Intelink domain (future efforts funded by the Horus program).

Metrics:

• Recall – number of instances annotated (class instances and property instances) divided by the number of instances that exist in the document expressed as a percentage
• Precision - number of instances correctly annotated (class instances and property instances) divided by the number of instances that were annotated expressed as a percentage
• Coverage - % of classes and properties in a given ontology that can be automatically annotated.

DAML Experiment:

The UBOT team plans to apply AeroDAML to support annotation generation for RDO/ONA domain. The UBOT team will work with other DAML teams to identify source documents and annotation details. Most of the technical goals listed above will make AeroDAML more suitable for supporting the DAML experiment in a variety of ways including simple fully automated annotation generation and more complex semi-automated annotation generation.

Next Logical Steps:

• Investigate the relationships between annotation metrics (recall, precision, and coverage) and the type of DAML application/query.

· Investigate algorithms for reconciling/mapping instances between annotated documents

DAML and UBOT Trial Application

Technical Goals:

Evaluate the DAML language and UBOT tools by applying them to the imagery and sensor domain. An agent-based system for satellite imagery analysis workflow is currently being developed. The UBOT ontology engineering tools are being used to develop and extend domain specific ontologies for this application. Specific tasks include:

• annotate collection source services with the DAML service ontology (DAML-S)
• annotate profiles for intelligence analysts with DAML
• integrate these DAML ontologies and annotations with workflow agents, broker agents and negotiation agents to coordinate activities

Metrics:

• Size/complexity of ontologies
• Degree of workflow automation

DAML Experiment:

This ongoing effort fits very well with the Joint Intelligence, Surveillance and Reconnaissance (JISR) part of RDO and the IXO mission. According to the RDO whitepaper: JISR “emphasizes collaboration between intelligence producers and users”. The UBOT team believes that including an adjacent domain to ONA in the DAML experiment will help demonstrate the benefits of DAML. “JISR is the front-end component of the ONA. JISR planning and execution are critical to the success of RDO. It is employed to gain information on adversary critical vulnerabilities and pinpoint potential causal links for ONA analysis.” The imagery workflow under development by the UBOT team addresses this specific area of JISR: “Multi-Intelligence Collaboration. Provides near real-time, collaborative tasking, processing, exploitation, and dissemination in national, theater, and tactical facilities regardless of geographic location.” The UBOT team is leveraging significant in-house JISR expertise. The UBOT team plans to continue working on a JISR demo in support of the DAML experiment and coordinate with other DAML teams working in the ONA domain.

Next Logical Steps:

• Expand the scope of the RDO/JISR DAML experiment to include other forms of intelligence sources and autonomous sensor-shooter capabilities.
• Work with Lockheed Martin Mission Systems to integrate DAML into their Joint Theater Logistics ACTD collaborative tools and asset visibility technology to support RDO objectives.