CSCI 6965 Advanced Semantic Web DistributedReasoningProject

From Semantic Portal Wiki

Contents 1 Discussion topics 2 Presentation outline 3 Related work 4 First Presentation Papers

Discussion topics

• query vs. reasoning
  • Shangguan: what is relationship between retrieval, query and reasoning?
  • Jesse: First intuition. I tend to think of querying as a matter of retrieval while reasoning is a matter of inferring or discovering new information. Clearly, there can be overlap, but I would think that not all queries involve reasoning, and not all reasoning involves queries. The two could be mixed, though.

• distributed vs. parallel
  • Jesse: First intuition. Distributed has to do with location, while parallel has to do with time. Parallel programming has to do with executing different parts at the same time, i.e. "in parallel." Distributed, on the other hand, means that parts of the process or data are in different locations. Parallel execution of an n-body problem on a cluster (for example) could be considered both parallel and distributed. It is parallel because different parts of the computation happen at the same time, and it is distributed because these computations happen on different nodes. However, a process simply running multiple threads of execution in parallel on the same machine is not distributed (if we consider a single "node" the smallest level of granularity), although it is parallel. A federated query over multiple, distributed data sources may not be parallel, but it is distributed in the sense that the data are located in different places. (Again, this is just a first intuition.)

Presentation outline

• Introduction -- Josh
  • definition of distributed reasoning, distributed vs. parallel reasoning

Shangguan 01:33, 22 February 2009 (EST): I think in addition to distributed and parallel reasoning, there are other reasoning mechanisms that can also be briefly stated here, e.g. p2p reasoning -- in fact, DRAGO is more like a p2p reasoning system. Actually, if we consider distributed reasoning as a goal, there exists various ways to implement it. So, maybe all of us could take some time to think about what these implementations may be.

• • distributed/semantically interlinked ontologies
  • distributed or parallel computation for reasoning tasks
    • partitioning of the rule base vs. the data set
      • time and space tradeoffs
    • communication and synchronization issues
    • implementation with web services
  • eager vs. lazy reasoning
  • tableau algorithms vs. rule-based reasoners

Shangguan 01:33, 22 February 2009 (EST): Besides the comparasion between distributed and parallel reasoning, maybe we can also compare both of them with normal reasoning/inferencing -- what's the difference, the extra work needed, what do we expect out of this distributed/parallel reasoning, etc.

• DRAGO / reasoning on distributed ontologies -- Shangguan
• Distributed reasoning using data partitioning -- Jesse
• Distributed reasoning using rule partitioning -- Giovanni
• Conclusion -- Josh

Related work

• P Adjiman, P Chatalic, F Goasdoue, M Rousset, L. Distributed Reasoning in a Peer-to-Peer Setting. ECAI, 2004
• Soma, R.; Prasanna, V.K., "Parallel Inferencing for OWL Knowledge Bases," Parallel Processing, 2008. ICPP '08. 37th International Conference on , vol., no., pp.75-82, 9-12 Sept. 2008 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=4625835&isnumber=4625815
• W3C. OWL-S: Semantic Markup for Web Services
• Luciano Serafini, Andrei Tamilin. DRAGO: Distributed Reasoning Architecture for the Semantic Web
• Distributed Reasoning with Modular Ontologies

First Presentation Papers

• Soma, R.; Prasanna, V.K., "Parallel Inferencing for OWL Knowledge Bases," Parallel Processing, 2008. ICPP '08. 37th International Conference on , vol., no., pp.75-82, 9-12 Sept. 2008 URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=4625835&isnumber=4625815
• Luciano Serafini, Andrei Tamilin. DRAGO: Distributed Reasoning Architecture for the Semantic Web