Semantic Web: a Working Symposium in Stanford

Part 2

The Semantic Web aims to provide new ways to define and link data on the Web, not just for display purposes, but for automation, integration and reuse across various applications. The full potential on the Web is going to be fulfilled with the ability to enable sophisticated Web services, for commerce and for exchange of high-value information. The Semantic Web project, as driven by W3C, is a mixture of theoretical and applied research work which takes advantage of various fields not connected until now. This project is designed to eventually provide standard solutions, independent of any proprietary implementation. Given the complexity of some of the applications being described, this task is not an obvious one, and this symposium was one important step on this path. The symposium provided a forum to present the state of the art in various technologies and approaches needed to make the Semantic Web a reality.

A symposium was held at Stanford from July 30 to August 1, 2001, to discuss the state of art of the Semantic Web. There were nearly 70 submissions, 250 participants. This workshop indicates that a new Semantic Web community is emerging. This article is the second part of a summarization of the highlights of this workshop.

Transactional Web versus Semantic Web

The relationship between the transactional web and the semantic web was the subject of an invited talk, by James Snell (IBM). The transactional web provides a technology foundation for the Semantic Web, while the Semantic Web provides a formal data model for Web Services. (The full presentation is available on the Web at http://www.semanticweb.org/SWWS/program/jamessnell.ppt)

Panel

A panel on "emerging semantics" was chaired by Vipul Kashyap. Ora Lassila (Nokia Research) presented his position as "optimistic and cynical". To the questions: is the Semantic Web going to happen?, his short answer was "maybe", and a longer answer was: "we have to make sure that certain key aspects are visible, such as ontologies and business models". For ontologies, the hard part is to get communities to agree. And for the business models, the question is: if the machines do the work, what do we do about advertising?". Dieter Fensel (Free University of Amsterdam) described ontologies as dynamic networks of meaning, and insists on the challenge of mastering the process of ontology evolution. Jim Hendler (University of Maryland, Baltimore County) describes the ways to reach consensus (bottom up or top down) and Clifford Behrens (Telcordia) distinguished two level: the molecular level (cultural model of domain) and the atomic level (classes for information management). He insisted on the fact that reaching an agreement relies on ontological notions that are not committed to a given tool.

Tutorials

Three tutorials have been offered, about ontology engineering, semantic B2B integration and E-services.

Ontology and Ontology Maintenance

Ontologies are an important player in the Semantic Web. Presentations went from specific ontologies to approaches to manage ontologies and "meta-ontologies".

One of the major problems facing actual ontologies is managing the changes in semantics. An example has been presented in a terminology in use by the Federal Drug Administration which led to the necessity of an "ontology of change" that could be generalized for the Semantic Web. Another example of ontology management was demonstrated with the products of VerticalNet called "Ontology Builder" and "Ontology Server" specifically targeted for e-commerce applications.

The question of how to connect various levels of ontologies was mentioned in a presentation of "OntoMap", a project aimed at facilitating the interaction between upper level ontologies and specialized, easy to use ontologies with the possibility of optimizing reuse of these resources. The process of building agreements among communities is not only technical, it also has social-cultural aspects which were described in a presentation about the making of semantic agreements on the Web. A tool called "the Schemer system" based on consensus analysis is a means to derive semantic knowledge from the information provided by subject matter experts.

Various ontologies used by library systems have been presented in order to derive requirements for structuring ontology systems.

The necessity of a higher level of modeling has been also the subject of a presentation on the use of UML for representing ontologies and domain knowledge on the Semantic Web. The paper proposed an extension of RDF allowing the identification of property-resource pairs in a model where usual reasoning based on a closed world doesn't work due to an environment where incomplete knowledge is a fact of life. XDD (XML Declarative Description) is a unified modeling language with well-defined declarative semantics. It provides mechanisms to express rules, conditional relationships, integrity constraints and ontology axioms. On a similar subject, an extension to RDFS, named "metamodeling architecture", can help resolve the layer problem.

DAML+OIL aims at facilitating the description of processes by software agents. An analysis of the difference between DAML+OIL and the original OIL (Ontology Interface Layer) shows that the modeling primitives are somewhat different. Another approach consists in using host formalisms (first order logic) to extend RDF semantics. An prototype built in Prolog, the RDF Schema Explorer, has been presented to illustrate this point.

An attempt to generalize RDF as a model of nested triple and lists of resources and triples enables to connect RDF with some of its proposed extensions. This might also lead to clarify the roles of reification and containers in RDF. Another proposal of a "Simple Ontology Definition Language" has been made with an application for a medical information service on the Web.

These open issues are among the ones that have led to the creation mid-August 2001 of the Web Ontology Working Group (WebONT W3C initiative). (see http://www.w3.org/2001/sw/WebOnt/ for more details).

The issues and priorities that have been discussed in this track can be summarized as a requirement for tools. Tools are needed for maintenance, versioning, collaboration, reasoning, merging, creation, validation, management of change.

Interoperability, Integration, Composition

DARPA is sponsoring an international collaboration called "Coalition Agents Experiment", allowing the construction of a coalition command support system, with agents groups into domains to reflect real-worldd organizational and national boundaries.

A geospatial application for visualizing U.S. election results has been presented with the objective to show the mapping between different XML representations and their conceptual models, the goal being to attain an overall framework for interoperation where maintenance problems are minimized.

The necessity of creating a framework to manage the market of digitalized multimedia has been emphasized. One presentation has insisted on intellectual property rights, another one has focused on the necessity to build an ontology for MPEG-7. This ontology should use a machine-understandable language, represented as an RDF schema.

The presence of DAML within the Semantic Web indicates an interest of the Artificial Intelligence community to be present on the Web. Traditional AI problems are therefore being imported within this arena. An example if that of when agents interact with other agents, they should use service descriptions that will let them know how to accomplish the other agents' goals. The problem of deciphering a service description is similar to the standard AI planning problem, but includes a need to reconcile contradictory ontologies and to rearrange data structures of a message-sending agent so they match the expectations of the recipient.

The mapping between the global ontology and the local ontologies involves querying. Query processing is strongly related to view-based query answering in data integration. This problem which is well known in the database world becomes now part of the tools needed to make ontologies work on the Semantic Web.

A simple intermediate conceptual model (ONION) describes the semantically relevant intersection of information resources with respect to a type of application. Application-dependent articulation rules capture the correspondence between concepts in different ontologies. This is a semi-automatic process that leads to the creation of an ontology algebra.

The interrelation between Topic Maps and RDF has been investigated in a presentation in order to define an interchangeable format for the exchange of knowledge on the Web. A possible first step consists in representing Topic Maps information as RDF information. The presentation is done by describing a mapping the topic map graph model to the RDF graph model.

When using knowledge coming from different sources, one of the most difficult challenges is to ensure correct understanding of the contents. A proposed methodology is based on analyzing notions of transformations from one language to another and of properties satisfied by transformations. Three steps were presented: (1) Define properties of transformations; (2) Express, in a form easily processed by machine; (3) construct by composition a proof of properties satisfied by compound transformations.

RDF can be used to build a programming language. RDF constitutes not only the means of expression, but also the subject matter of programs.

RuleML is a rule markup language proposed for the Semantic Web. RuleML implementations are made via XSLT and rule engines. A review was made of the specifications of RuleML leading to the expression of requirements for a future version of RuleML.

Wilbur is a representation of the RDF data model as a frame-based representation system. It is also aimed at a programming system.

Web Services and Applications

A number of Web services application scenarios were presented within the third track.

An ontology of services is considered necessary It will contain the structure of the ontology, the service profile for advertising services, and the process model for the detailed description of the operation of services.

The major goal of these Web Services ontologies is to help users quickly the services they need.

As en example, advanced matchmaking services require a new set of rich and flexible metadata not currently supported by the available industry standard frameworks such as UDDI and ebXML, in use for e-commerce. A prototype containing matchmaking features has been presented, in relation with RDF and DAML.

Authoring tools for generated ontologically encoded descriptions are needed. The "Briefing Associate" augments Microsoft's PowerPoint to support the authoring of semantically grounded briefings.

ITTALKS is an automatic and intelligent notification of information technology talks using DAML to provide online services for humans as well as agents.

The domain of e-learning has its own requirements, which have been addressed in a presentation focusing on Open Learning Repositories and Metadata Modeling. This approach enables its users to express information about the learning objects contained in the repository. It makes also possible the expression of relationships between these learning objects. RDF/RDFS and O-Telos modeling are used.

"CREAM" (Creating RElational, Annotation-based, Metadata) is a framework for an annotation environment. relational metadata, i.e. metadata that comprises class instances and relationship instances. The instances are based on a domain ontology.

A comprehensive solution to Web site and Web portal creation, relying on a model-driven, ontology-based, web site management, dubbed "OntoWebber", contains three different aspects: (1) explicit modeling of different aspects of Web sites, (2) use of ontologies as the foundation for Web portal design (3) semi-structured data technology for data integration and Web site modeling.

Building a structured index of a web site is done through use of ontologies and natural language techniques for information retrieval on the Internet. A technique has been presented using improved natural language techniques to extract well-formed terms taking into account HTML markers.

WebML (Web Modeling Model) is semantic model for WebML hypertexts by means of state charts.

Birds of the Feather Sessions

Several Birds of the Feather sessions took place at the end of the workshop.

One of them was about business models for the Semantic Web. Business-wise, the Semantic Web is only at its very beginnings. Hence, there are many aspects considered still unclear: there seems to be a need for an emerging "business ecosystem" around the Semantic Web, where ontologies occupy a place of choice. A detailed summary of the discussions can be found at http://business.semanticweb.org/swws-bof-discussion.html.

Douglas Engelbart was among the participants to this workshop. One session was devoted to his "bootstrap philosophy" which was summarized as such: "Eat you own dog food: build tools, publish them often on the Web, this will build the community."

Part 1 of Mr. Biezunski's report on the Semantic Web Symposium in Stanford appeared in the October Issue of XML Files.

Return to TOC