|Title / Titel
||High-Precision Service Retrieval
||Abstract (PDF, 14 KB)
|Summary / Zusammenfassung
||Increasingly, on-line repositories such as the World WideWeb are being called upon to provide access not just to documents thatcollectuseful information, but also to servicesthat describe or even provideuseful behavior. Potential examplesof such services abound:
As the number of such services increases it will becomeincreasingly important to provide tools to quickly find the servicesthey needwhile minimizing the burden for those who wish to list their serviceswiththese search engines. Current service retrieval technology is, however,notoriously prone to low precision and poor recall as well asoftentimes toocostly to use.
- Software applicationssuch as web services that can be invoked remotely by people orsoftware. E.g., www.salcentral.com.
- Software components thatcan be downloaded for use when creating a new application. E.g.,www.mibsoftware.com, www.compoze.com.
- Best practice repositoriesthat describe how to achieve some goal. E.g., process.mit.edu/eph/,www.bmpcoe.com.
- Individuals or organizations,who can perform particular functions,E.g., guru.com, elance.com, andfreeagent.com.
Our research in this area so far has proposed a novelservice retrieval approach based on the sophisticated use of processontologies. This approach uses a semantically rich description of aservice’s process semantics to improve query performance. While ourpreliminaryresults are promising, many important challenges remain. At the currentstatethe project provides a good basis for a deeper exploration andevaluation ofprocess-based service retrieval. In this project we, therefore, proposeto:
Furthermore, as process-based annotations and ontologies cancomplicate the use of service retrieval technologies, theoverarchingquestion of the proposal is to explore (and possibly improve on) thecost/benefit tradeoff between rich service descriptions and ease of useforboth the service providers and the service seeker. To that end, thestudyplans to:
- Further evaluate if and howprocess-based service retrieval improves retrieval performance(precision and recall) using realistic/real-world settings.
- Find means to further improveretrieval performance.
Theimpact of this project is twofold. On thepractical side, it is an enabler for the retrieval of serviceproviders, whichis a critical element for the large-scale use of web-services,multi-agentsystems, and could significantly simplify the discovery of partners foroutsourcingarrangements enabling highly dynamic (virtual) enterprises. On thescientificside, the project contributes to the body of knowledge developed in thesemantic web and in areas such as multi agent systems. It also providesa basison which future distributed computing applications could rely to enablethedynamic construction of applications.
- Experimentallyexplore thecost/benefit tradeoff between semantic markup and service retrieval.
- Simplifyuser tasks such as querying for and indexing services to improve thebefits of using process-based retrieval.
|Publications / Publikationen
|Keywords / Suchbegriffe
||process models, service discovery, resource discovery, semantic web, matchmaking
|Project leadership and contacts /
Projektleitung und Kontakte
|Prof. A. Bernstein, PhD (Project Leader)
|Other links to external web pages
|Funding source(s) /
|SNF (Personen- und Projektförderung)
Support from US funding agencies for US research partners
|In collaboration with /
In Zusammenarbeit mit
|Mark Klein, Center for Coordination Science, MIT
|Duration of Project / Projektdauer
||Mar 2004 to Mar 2006