Tools for DAML-Based Services, Document Templates, and Query-Answering
Intent of Work (IOW) for FY03 (October 2002 – September 2003)
Sponsor: Department of the Air Force
Prof. Richard Fikes, Principal Investigator
Dr. Deborah McGuinness, Technical Program Manager and contact
Dr. Sheila McIlraith, Web Services Technical Lead
Knowledge Systems Laboratory
Computer Science Department
KSL continues to play a leadership role in DAML language design and extended environment specification, committee work, and education work. This is critical to the program to provide both a sound foundation for the DAML language and the necessary supporting environment for use. In the next year KSL will provide:
* Joint US/EU ad hoc Agent Markup Language Committee participation on DAML language evolution. McGuinness and Fikes will continue their roles on the committee, including Fikes continuing to lead the DQL query-answering language and protocol development and McGuinness leading the explanation focus.
* World Wide Web consortium Semantic Web Activity Web Ontology working group participation. McGuinness continues to represent Stanford and the DAML community. She leads the OWL Lite and Feature Synopsis specification and document efforts and is a co-editor of the OWL reference specification and OWL Guide. She also leads the example wine agent work that is used to demonstrate a common use case of markup languages on the web.
*Continued maintenance of the DAML Axiomatic Semantics as the language is extended and the semantics clarified. We will author an axiomatic specification for the Ontology Web Language (OWL) being developed by the W3C WebOnt Committee. We will continue our partnership with Waldinger at SRI, who is validating the axioms using automated reasoners and will co-author the OWL axioms.
The work provides the foundation of the DAML program providing specification of the language, specification of the core extensions such as the query language, the rule language, and the basic use cases, and the support infrastructure for communication and education.
Measures of progress are: the use of DAML both within the DAML project as well as in the outside world, the acceptance and use of the W3C web ontology language, and the use of the axioms. Some of the education and documentation will support the DAML experiment.
KSL has played a technical leadership role in the development of DAML-S through participation in the DAML Services Coalition. In the following year, the Coalition will transition to the Joint US/EU Semantic Web Services Committee. As a founding member and member of the Steering Committee, McIlraith will be working on the organization and initiation of various subcommittees. As a technical member of the Languages Subcommittee, she will work towards the redesign and deployment of DAML-S or its successor. In the interim, she will continue to contribute to the development of DAML-S.
Services can be characterized as information-gathering and/or world-altering. KSL will develop ontologies for information-providing services as part of DAML-S or its successor. These will support query answering from knowledge on the semantic web (see below).
KSL will complete development of prototype components for a DAML-S Web service annotation tool. This tool will be used by service providers to describe the capabilities of their services in DAML-S. It will facilitate the markup of both new and existing Web services. KSL had hoped to work with Booz Allen to realize the full development and testing of this tool. We will seek out an appropriate alternative partner for this development in the coming year.
KSL will continue its development of technology for DAML-enabled Web service composition and interoperation. Web standards defined by industry efforts will shape the evolution of Semantic Web services. In contrast to the top-down effort in DAML-S and associated agent technology developed by the DAML program, we will take a bottom-up approach, taking the leading industry candidate for Web service business process modeling, BPEL4WS, and its associated computational machinery, BPWS4J, and augment them with Semantic Web technology, in particular OWL, DQL and JTP, to perform automated Web service discovery and semantic translation. Our objective is two-fold: to demonstrate to the mainstream Web Services community, the value-added from semantic Web technology, and to examine how DAML+OIL and OWL can integrate with emerging industry standards in business process modeling.
KSL will continue its development of technology for DAML-enabled automated Web service composition, with a focus on information integration, user preferences, and failure recovery.
KSL will continue its development of a methodology and supporting technology for query-answering from knowledge expressed in semantic markup on distributed Web sites. In addition to addressing the standard issues about how to do effective reasoning with knowledge expressed in a highly expressive formal logic representation language (e.g., DAML), our work is addressing the significant additional issues raised by the knowledge being represented in markup on user Web site and using terminology defined in (typically large) ontologies resident on (typically multiple) other Web sites. For example, reasoners are not going to be able to do anything significant working directly from the markup, and one doesn't want to pay the overhead of loading knowledge into a reasoner’s internal memory every time one wants to get an answer to a query.
The methodology we are developing is grounded in the use case goal of associating with a Web site containing DAML markup an agent that is an expert on the knowledge represented in the DAML markup on the pages of that Web site and that provides a set of information services based on that knowledge. We are developing an API and architecture for such agents to support providing a DQL (DAML Query Language) query answering service using the markup on the site as its knowledge base and some or all of the content of that knowledge base in various forms such as a set of RDF statements, a KIF logical theory, an HTML document suitable for presentation on various output media, etc. We are exploring the notion of extending that idea to say that the *only* thing one encounters when accessing DAML markup is such an agent, where one of the services of the agent is to provide the knowledge represented there in text as a DAML element. The agent will act like a knowledge server for the knowledge represented in DAML markup in that it will keep the content of the markup loaded and include a query answering and constraint checking service.
The supporting technology we are developing for this task includes (1) a query-answering reasoner called JTP implemented in JAVA suitable for use as a primary component in an agent that is an expert on the markup on a given Semantic Web site, and (2) a query language and supporting query-answering protocol for OWL, which will be called OQL.
JTP contains a general-purpose reasoner integrated with a collection of special-purpose reasoners that are suited to DAML+OIL and is being augmented incrementally to provide expert reasoning in specific task domains. We have successfully tested and demonstrated JTP’s ability to answer queries from small to medium sized DAML+OIL knowledge bases and have delivered an initial version of the system to the DAML integrators (BBN). During the coming year, we plan to upgrade the system to be usable with (1) larger DAML+OIL knowledge bases, (2) knowledge bases that use ontologies distributed over multiple Web sites, (3) OWL knowledge bases, and (4) the anticipated extended expressive power of the full DAML language. We will complete implementations of both (1) a human Web-based user interface that will allow JTP to be used as a query-answering agent for a DAML Web site (as described above) and (2) the DQL query-answering API for use as a DQL server by computer-based agents. We will also (1) continue our participation in the development of an ontology for representing time dependent knowledge being led by Ed Hovy, (2) develop a special-purpose reasoner for answering queries from knowledge represented using that time ontology, and (3) develop a general methodology for associating with an ontology special-purpose reasoners for knowledge represented using the vocabulary specified in the ontology.
OQL (OWL Query Language) will be a formal agent-to-agent language and protocol for query-answering using Semantic Web knowledge represented in OWL that supports automated reasoning methods to derive query answers. OQL will be based on DQL, which was developed by the DAML Joint Committee led by KSL. We will submit the completed OQL specification to the W3C WebOnt Committee as a candidate standard interface for agent-to-agent query-answering from OWL knowledge bases. In addition, we will continue to lead the evolution of the design of DQL based on feedback and on the anticipated expansion of the DAML language.
Evaluation of both DQL and JTP can be based on their ability to effectively support the query-answering capabilities that will be needed in the upcoming DAML demonstration and experiment. Evaluation results would be in terms of specific kinds of queries (perhaps specified as parameterized queries as in the DARPA HPKB program) asked of specific knowledge bases distributed over a specific number of sites.
The ability to automatically reason with knowledge expressed on Web pages is a central notion in the Semantic Web vision and can be expected to play a significant role in the upcoming DAML demonstration and experiment. A reasoning capability enables answers to be determined for queries from Web sites that do not explicitly state those answers. The reasoner infers the implicit answers by using the content of the Web site and of the ontologies used by the Web site. In addition, a reasoning capability can determine whether the information expressed in semantic markup on a Web site is logically consistent with the ontologies referenced by the Web site. Such reasoning is a core capability of semantic markup that distinguishes its use from the use of XML and HTML.
DQL, OWL, and the JTP reasoning technology we are developing provide the key elements of a query-answering facility that performs deductive retrieval from knowledge represented on the Semantic Web and a diagnostic facility that checks the logical consistency of such knowledge. The query-answering facility can be used by any service or tool that is retrieving information from Semantic Web knowledge bases, and the diagnostic facility can be used by any service or tool that is supporting the authoring of Semantic Web ontologies and knowledge.
In the upcoming DAML demonstration and experiment, we will provide JTP as a component tool and support its use as both a Web site query-answering agent and as a diagnostic tool for use in ontology development.
KSL will continue its work on ontology environment specification and tools for DAML+OIL and OWL. In addition to the effort focused directly on query answering, we have begun work on an interoperable explanation effort to facilitate trust of DAML+OIL and OWL applications. We also continue to enhance the KSL tool suite to support DAML+OIL and OWL needs where appropriate.
*Inference Web. Inference Web provides an infrastructure for supporting interoperable explanations of answers from semantic web applications. It is composed of a portable proof specification, a browser for displaying justifications, a registry of reasoners, inference rules, and ontologies (with meta information), and an API for interacting with the components. We have published an initial portable proof design (in DAML+OIL) that is to be used for proof sharing. We are obtaining feedback from reasoner authors and users in the RKF, DAML, Ultralog, AQUAINT, and KD-D programs for adequacy of the design. KSL will continue work on the portable proof specification and provide an implementation of Inference Web using the design. We have also begun work on a browser that displays information provided in the portable proof format. KSL will provide a distributable implementation of the proof browser. We have also begun work on the registry containing an initial specification for inference engines, inference rules, and ontology meta information. KSL will provide a prototype registration implementation and service. More information on Inference Web can be obtained from http://www.ksl.stanford.edu/software/iw/ .
* The KSL JTP reasoner is being developed and enhanced for DAML and OWL. It is largely driven by the query answering work and is elaborated on in section III. It provides a portion of the ontology development environment infrastructure. It is a hybrid reasoning architecture and may be used to incorporate other special purpose reasoners such as FACT or our time reasoner. It is being enhanced to provide explainable output using our Inference Web infrastructure using DAML+OIL and OWL.
* KSL and colleagues continue to utilize the Chimaera Ontology Evolution Environment tool aimed at supporting ontology merging, analysis, and transition. Chimaera has been and is being used in to diagnose the W3C guide knowledge bases. Chimaera will be upgraded to support OWL reading and writing.
Ontology development environment work is critical and leverageable by the DAML experiment. The JTP reasoner and the Inference Web environment may be used to generate consequences of DAML statements that are explainable. Additionally the KSL Chimaera ontology evolution environment may be used to analyze and merge DAML+OIL/OWL ontologies.
Measures of success with respect to the DAML experiment are use and use without consulting of the tools. Measures of success of the work in general is use of the tools inside and outside the DAML community.