International Semantic Web Working Symposium
SWWS

sponsored by the
National Science Foundation, Information and Data Management Program

Stanford University, California, USA
July 30 - August 1, 2001

Report

Integration, Interoperation and Composition Track

Vipul Kashyap
Applied Research, Telcordia Technologies
kashyap@research.telcordia.com

Abstract

This is a report on the proceedings of the track on Integration, Interoperation and Composition as a part of the 1st International Semantic Web Working Symposium, held in Stanford from July 30 - August 1, 2001. The track description and organization is first discussed. Then we present a brief discussion on the issues discussed and the missing gaps identified for enabling the Semantic Web. Next we shall discuss the research agenda and identify the priorities for academia and industry.

Track Description and Organization

Interoperation and Integration can be defined by broadening the current definitions used in the federated database literature. Interoperation may be defined as a loose coupling across information sources, semantic metadata descriptions and ontologies. Typically an external agent or system is responsible for decomposition of an information request and mapping the decomposed pieces onto the given targets. Integration on the other hand refers to a tight coupling and suggests that the given information sources, semantic metadata descriptions are somehow coalesced together and can be queried as a single unit. Composition on the other hand refers to the act of combining pieces of information at a level granularity smaller than that involved in integration and interoperation. Examples of pieces are: constraints, policies, vocabularies and language transformations. An interesting case of composition is when you may want to compose proofs of correctness to prove higher level security and transaction policies.

Based on the above definition the track was (approximately) divided into four broad sessions involving 12 presentations.

• Session 1: This section focused on applications demonstrating the value of semantic integration, interoperability and composition. This session had 3 presentations
• Session 2: This session primarily focused on ontology integration issues in multimedia and agent based systems. This session had 2 presentations
• Session 3: This session focused on frameworks for ontology integration and semantic interoperability. This session had 4 presentations.
• Session 4: This session focused on "semantic web programming", i.e. issues related to gluing together web sites based on the semantics of information. This included markup and representation languages and adaptation of standard programming languages for this task. This session had 3 presentations.

Issues

The central observation of all the speakers presenting in the track can be summed up as: It is crucial for the interoperability layer to migrate from the syntactic to the semantic!

Semantic Interoperability (in the general sense including integration and composition) could be implemented at different levels, and there were proposals to organize them approaches in a layered manner. This is illustrated in the figure below and seeks to differentiate between data networking and "semantic" networking.

Building on the success of the data networking and middleware communities, the above picture tries to relate organize and relate the semantic web efforts along multiple layers, some of whom are described below:

• Object Interoperability: This is the layer at which the current middleware products are aimed in the industry. However these objects are primarily defined as containers for software and for streamlining the software development process. The CORBA, EJB object models are examples of standards at this layer.
• Meta-Model Interoperability: This is the layer at which the cross-over from the "data" space to the "knowledge" space takes place. The objects here are viewed as containers of knowledge to be fleshed out by upper layers. The OKBC and RDF(S) core models are examples of standards at this layer.
• Ontology Interoperability: This is the layer where ontologies, schemas and classifications are built upon common underlying standardized meta-models. The ability to use different ontologies to specify and query information constitutes interoperability at this layer.
• Meta-Data (View/Query) Interoperability: Semantic metadata descriptions can be constructed from one or more underlying ontologies. Issues at this layer would be to decompose information requests into those supported by the individual semantic metadata descriptions corresponding to the information sources.

The ability to organize semantic web research along these layers helps us organize the work require to build out the underlying infrastructure of the semantic web. The issues that arise are: development of standards and industry wide APIs at each of the layers. Building up semantic-web specific functions such as semantic routings, "semantic" content delivery networks. Specification of further application layers may also be required.

One of the most important topics on which semantic interoperability depends is the problem of ontology interoperation, which was discussed by a large number of speakers. Some topic that were discussed were: integration between RDF, Frame based and Object Oriented data (meta-model reconciliation), merging and exchange of data between different RDF models, and the ability to specify mappings across terms in different ontologies. Some of the issues that were discussed in the context of the latter topic were the reconciliation of various types of heterogeneities and techniques for specifying articulation rules and correspondence across ontologies.

Languages for representation of ontology transformations/mappings and for "semantic" web programming were also discussed. Approaches to represent computations in RDF were presented and their similarity to the notion of web services was noted. Representations of active/re-active rules and action using RDF, and integration of RDF/DAML+OIL with a programming language were two diverse approaches discussed. An interesting paper on proving properties of transformations based on composition of the transformations (and their corresponding proofs) was also presented.

Some interesting presentations dealt with the issue of semantic web in the presence of multimedia data. The MPEG-7 standard was discussed and the need to specify the semantics of the MPEG-7 metadata was recognized. The ability to compose semantics of document components in the presence of spatio-temporal constraints was identified as a crucial requirement for multimedia semantics. The need to link and interoperate with ontologies from other domains was also recognized. Digital Rights Management was especially an important topic from the perspective of multimedia data and there were discussions related to rigorous specification and enforcement of digital rights associated with a multimedia document. There was a suggestion from a participant that ongoing and previous work in digital libraries should be leveraged for the semantic web.


Missing Gaps

Whereas there were a lot of issues relating to the Semantic Web covered by the speakers in the track some critical problems that need to be addressed had not been covered. Some of the problems that comprise the missing gap are:

• Ontology Impedance: Ontology impedance may be defined as the semantic mismatch between two or more ontologies that are being merged. There was widespread acceptance of the fact that, when merging ontologies, we will have to deal with ontology impedance. Work needs to be done to estimate the consequent loss of information that results from this impedance.
• Scalability/Performance: Issues related to scalability of web servers serving semantic web content is a critical issue on which the future semantic web depends. Work is needed to come up with techniques that exploit "semantics" to design better caching techniques, e.g., semantic content distribution networks. There is also a need for metrics and measurements to evaluate how well algorithms for the semantic web will perform and scale.
• Dynamic Ontologies: A fundamental but flawed assumption being made by all the speakers was that ontologies are static in nature. The techniques presented were based on the assumption. Real world ontologies are likely to be dynamic and evolve over time and algorithms and techniques for the semantic web need to be adapted to account for this possibility.
• Semantic Metadata Extraction: Two crucial factors that will determine the success of the semantic web are: the ease and cost of developing and maintaining ontologies, mappings and articulation rules; and the ease of constructing semantic annotations. Tools that drive the extraction process based on text processing and NLP techniques (most of the data on the web is textual) are important.

Research Agenda

The research agenda for the Semantic Web can be divided along the following lines. Industry research typically needs to focus on shorter term, applied research with emphasis on proof of concept prototypes that can be productized if there is commercial potential. Academia perform longer term fundamental research, since they don't face commercial or business pressures. We now try to outline the research agenda for industry and the academia in the context of the semantic web.

Some research priorities for the industry are as follows:

• Scalability/Performance: This is likely to be one of the topmost priorities of the industry. As new semantic web technologies emerge, research needs to be done to ensuring the scalability of these technologies.
• Multimedia Data: The industry is likely to be interested in semantic annotations to various types of multimedia data, such as images, video. Current interest is centered around specific information domains such as GIS, Medical Images and News Videos.
• Ontology Integration/Interoperation: The applications of ontology integration/interoperation techniques that the Industry is likely to be interested in is catalog integration and business process modeling and integration. There is a need for definition and interoperation across multiple process based ontologies.
• Digital Rights Management: Applications related to digital rights management especially those related to formal semantic specifications of digital rights might be important for industries as they try to protect their intellectual property over the web.

Some research priorities for academia and long-term projects undertaken by some industrial R&D Labs are as follows:

• Ontology Integration/Interoperation: Given that ontology mismatch is a certainty on the Semantic Web, research is needed to categorize the different types of mismatch and metrics to measure the amount of mismatch. Metrics for mismatch can be adapted to estimate loss of information.
• Languages for Ontologies: Representational languages with specific properties might be required to represent ontologies and transformations across ontologies represented using different languages. Correctness proofs about these transformations are an important area of research.
• Semantics of Multimedia Data: Multimedia data, typically consists of components that might be of different digital media. The challenge here is to "compose" the component semantics to determine the semantics associated with the multimedia document. Spatio-temporal constraints play an important role for capturing semantics of multimedia data and need to be explored.
• Tractable areas of high complexity problems: It is well known the inference and proof procedures for complex logics, likely to play an important role for capturing semantics, are computationally expensive and intractable. Research is needed into tractable techniques such as Markov processes for inferences and processing.
• Expression of "shared meaning" in the social sense: Ontologies and shared knowledge develop as a result of social and consensual processes. Current semantic models assume consensus and focus on formal representations of semantics. Research is needed to come up with semantic models that capture both the formal and the consensual nature of meaning.
• Semantic Digital Rights Management: Concepts like trust, credibility, permissions and obligations are likely to become important in the future semantic web. Research on proof procedures and modal logics to establish these concepts.
• Cognitively enabled Ontology Authoring: Current ontology development tools are evolving to incorporate usability principles for ontology authoring. There is a need to research into human cognitive abilities and define tools that take advantage of those abilities.

Industry and academia both seek funding from Government agencies such as DARPA, NSF, etc. Industry in particular seeks funding for long-term projects that do not have immediate commercial viability. It is a moot point whether the semantic web will be the next "internet" to be funded by DARPA. Government funding may turn out to be an exciting way to promote industry academia collaboration, as seen in the CoAx project by Allsopp, et. al.

Conclusion

In conclusion, the Interoperability, Integration and Composition track had an interesting collection of papers, but crucial problems important for the enabling of the semantic web were either missing or not dealt with. Some open difficult problems were alluded to but not explicitly dealt with. In general, the track had a very strong academic flavor and attempts should be made to involve the industry in a more significant manner.

.