Description of work         Excepted Results      Workpackage List         

Description of work

The structure of the project workplan runs as follow. It first recapitulates on the motivation and the underlying techniques of SWAP. It then derives concrete tasks that must be solved for P2P and ontology-based knowledge management. In order to remain open to new ideas, it provides a generalized framework abstracting from the concrete tasks. This framework will be used as a first guide that serves to explore the terrain of P2P and Semantic Web, as a starting point for conceptual development as well as a baseline for future ideas on methods and tools for Semantic Web and Peer-to-Peer.

The project will combine/harness the power of P2P and ontology-based Knowledge Management. We will demonstrate solutions that are capable of identifying/discovering semantics, as it emerges from actual use. It will let people cooperate with each other in a peer to peer fashion. The end result will be a scalable Semantic-based P2P environment where participants can seamlessly share and update knowledge across dynamic communities of practice and other dynamic groups of participants with overlapping interests.

The project will address these issues along five main lines of action:

Expected Results

Tasks of SWAP will, in particular, account for four challenges that can be derived from some of the major differences between ontologies in a P2P environment and from ontologies in a centralized client-server environment, which is the usual setting in current applications, viz.:

• Peer selection service: In order to receive the right answers without flooding the peer network with queries one must ask the "right" peers. We will investigate Ontology-based peer selection mechanisms and implement a corresponding tool (called P-S).
• Variation of ontologies: Different peers will use different, though overlapping ontologies. Alignment, mapping and visualization tools (P-Al, P-Map, P-Viz, resp.) will have to cope with different ontologies, even though no alignments are explicitly specified. Some of the alignments and the mappings may be found by analysis of peer knowledge using methods of the just emerging field of Emergent Semantics (e.g. same file categorized to different concepts indicates alignment). Another tool, P-Oc, will scrape ontologies from legacy information (e.g. folder structures).
• Lack of ontological precision: Ontologies will be produced from various user interactions, like classifications into folders or usage of metadata. Ontology definitions will be imprecise and "sloppy" ontologies will be the norm rather than the exception. An inference engine for these ontologies, P-I, must be able to ask and answer queries to peers in a robust, scalable, often locally contained, manner.
• Ontological drift: In a P2P environment, one cannot expect any maintenance to happen on the ontologies (in fact, users will often not know what is in the ontologies on their machine). As a result, we must design mechanisms that allow the ontologies to update themselves, in order to cope with ontological drift. Based on the queries and answers elsewhere in the P2P network, ontologies will have to adjust their own definitions accordingly (using P-Up).

In order to evaluate our approach, SWAP will provide an ontology-based P2P platform that integrates these tools and that serves as foundation in two case studies:

First, in the Knowledge Management Case Study we will investigate the work of investment analysts at Dresdner Bank. The case study will survey their use of current central knowledge management technology that the investment analysts already use. Furthermore, it will chart the possibilities for P2P based solutions of knowledge sharing that have not been undertaken, yet, because the centralized approach currently in place would involve too much overhead for administration and analysts alike. The P2P based solution will be installed and test-driven in order to come up with a realistic estimation of the work that might be saved (or wasted) by the P2P system.

Second, in the Virtual enterprise case study, contractor IBIT provides telematics expertise for purposes such as tourism and SME support in its geographical region. There is a tremendous need for better knowledge sharing between many of the small tourism SMEs (e.g. tour guides, tour operators, etc.). SWAP will investigate how P2P- and ontology-based solutions may make knowledge sharing inexpensive, yet effective in such virtual enterprises.

Concrete results of SWAP will be

Measurable evaluation criteria for the case studies will in particular include:

• How many PCs could be reasonably connected? Can one give an estimation for an upper limit of computers that can be connected with SWAP? If yes, what is this number?
• How large are the ontologies used in the case studies?
• How does the quality and statistics of the ontologies in the case studies compare to conventional ones (e.g. compared to http://www.daml.org/ontologies/)?
• How much manual effort was needed to turn the single ontologies productive?
• How much overlap was between the different ontologies such that SWAP could work in the case studies (e.g. in terms of overlap between ontologies at different levels)?

The case studies will include particular efforts to measure these criteria (and further ones to be determined by ongoing development)

Thus, SWAP aims at resulting systems and experiences that can be seen as a giant leap towards the Semantic Web where both structure and content are truly distributed and (semi-)automatically administered. The research groups and companies in this project are pioneers in the research, development and commercialization of ontologies and semantic web solutions. They have all the capabilities needed to combine Semantic Web and P2P – adding a distinctive European dimension to mostly US efforts like Gnutella, Napster, or Jxta. Europe is uniquely positioned to continue the tradition of a bold mover in the Semantic Web and to plough the grounds in terms of academic and commercial research and development for the not-so-future key technologies.

Workpackages List.