Agent-Based Question Answering System. This project aims to improve our AQUA system – a question answering system which was developed as centralized version in an earlier version. Although, this project has been carried out as part/time project.
- MAOM-QA Multi-Agent Ontology Mapping for Question Answering. In this project we have carried out research on the Ontology mapping problem (also called ontology alignment). Our approach to ontology mapping is an agent based approach where we have several agents performing mappings and then we combine the evidences found by each of them using Dempster-Shafer Theory. This solution is generic as the system does not need to learn mappings in advance like other approaches which use Machine Learning techniques. Our mapping algorithms get evidences from different sources like wordnet and background knowledge (specific to the domain’s ontology). Furthermore, our approach deals with uncertainty in mappings whilst other approaches only handle two valued logic. Note that we (Dr. Maria Vargas-Vera and Miklos Nagy) were pioneers on the use of the Dempster-Shafer Theory coupled with a Fuzzy Voting Model on the Ontology Mapping problem. Our work can be followed from 2004 when my team started solving the ontology alignment problem using the Dempster-Shafer approach and assessed the main limitations on the original rule of combination (refer to the two full publications appeared at Conferences in 2005). Some years later (2006 and 2007), other researchers followed our approach to Ontology Mapping, in particular, researchers from the Ontology Mapping and Semantic Web Community. This has helped to broaden the general knowledge on this area. As a second phase in our original proposal, we included into Dempster-Shafer rule of combination a Fuzzy Voting Model to resolve the problem of “contradictory evidences” which are ignored when using Dempster-Shafer rule of combination. The Dempster-Shafer rule strongly emphasises the agreement between multiple sources and ignores all the conflicting evidence through a normalization factor. One of the outputs of this project is the DSSim System (short for Similarity based on Dempster-Shafer) which is written in Java. We decided to evaluate our approach (embedded in a system called DSSim ) using large and real ontologies. Therefore, we have participated in the Ontology Alignment Evaluation during several years. This participation helped to tune our algorithms over several iterations. Then in the Evaluation 2008, due a excellent performance of our system, the DSSim team was invited for a oral presentation at OM-2008, Germany 2008. Below are details of our participation in the Ontology Alignment Evaluation from 2006 until 2009.
- We had participated in the Ontology Aligment Evaluation Initiative 2006 with our DSSim system (v0.1). However, we only participated on the benchmark track.
- We had participated in the Ontology Alignment Evaluation Initiative OAEI 2007 with our DSSim system v0.2. The Evaluation OAEI-2007, provided problems on several tracks but we only participated on benchmark, anatomy, food, environment and library tracks.
- We have participated in the Ontology Alignment Evaluation Initiative 2008 with our system DSSim (v0.3), together with Department of Information Systems, Poznan University of Economics, Poland. The OAEI-2008, provided problems on several tracks. This year we participated on all given tracks for the evaluation 2008: benchmark, anatomy, fao, directory, mldirectory, library, very large crosslingual resources and conference. For the year 2008, we included two new components “Compound Nouns” and “Fuzzy Voting Model” into our DSSim (v0.3) system. DSSim obtained exceptional results in the OAEI-2008 Evaluation. Read more.
- DSSim made the news headlines in October 2008.
- More news on DSSim in the year 2008
- Also, we have participated in the Ontology Alignment Evaluation Initiative 2009 with our system DSSim (v0.4), together with Department of Information Systems, Poznan University of Economics, Poland. This year (2009) we participated in 7 tracks for the evaluation 2009: benchmark, anatomy, directory, ars, iimb, very large crosslingual resources and conference. Results can be found on paper Results of the Ontology Alignment Initiative 2009. by Jerome Euzenat et al. presented at OM-2009; collocated with the ISWC 2009, Washington DC, USA, October 25 2009. A summary of our participation in 2009 can be found at DSSim’s participation summary.
- Further information on DSSim:
Current work on this project involves applications of DSSim in Configuration problems.
Great minds discuss ideas; average minds discuss events; small minds discuss people. Eleanor Roosevelt