Ontology mapping with intelligent agents on the Semantic Web : the theory and practice of agent belief and consenus building

Nagy, Miklos (2012). Ontology mapping with intelligent agents on the Semantic Web : the theory and practice of agent belief and consenus building. PhD thesis The Open University.

DOI: https://doi.org/10.21954/ou.ro.0000faa5


Many practical scenarios in the Semantic Web require domain independent semantic data integration between heterogeneous sources. Therefore, ontology mapping is considered as an important problem that needs to be resolved. For example, in 2008 Oracle1 predicted that the worldwide data integration software market would exceed US$3 billion by 2012. In response to this need, this research investigates how a domain independent and scalable mapping solution can be devised using a multi-agent architecture. The proposed solution uses uncertain reasoning and conflict resolution, in order to establish high quality mapping results across various domains.

The review of the literature on traditional database and semantic data integration reveals the limitations of existing technology, in areas such as domain independence, reasoning with incomplete and uncertain information, conflict resolution, and scalability. In response to these limitations, I present a multi-agent solution that uses uncertain reasoning with belief functions over similarities. The proposed approach is not based on any heuristics or domain specific rules but instead aims to assign a meaning (interpreted by the software agents) to the entities that are present in the different ontologies, independently of their representation. The research investigates how domain independence of an ontology mapping agent can be achieved in order to formulate beliefs in the similarity of the terms between mapping pairs. I propose to manage uncertainty, a focal issue that is inherent to data interpretation, through Dempster-Shafer belief functions. Furthermore, the research explores how conflicts in belief functions can be treated with voting mechanisms that establish trust in agent beliefs, while rejecting beliefs that cannot be trusted in a certain situation. Additionally, an approach to optimise belief aggregation in a multi-agent environment is presented that makes my solution viable for domains of all sizes.

A series of algorithms have been developed in order to validate the proposed theory on how to approach domain independence in the ontology mapping process. These algorithms form the core part of the DSSirn ontology mapping system that can use different ontology representations based on Semantic Web standards, such as OWL. An empirical evaluation of the DSSim system is presented, which is based on an international benchmarking initiative. DSSim has participated in the yearly Ontology Alignment Evaluation Initiative (OAEI) evaluations from 2006 up to 2009 and has been tested on 22 tracks. Given its wide range of domain coverage, the OAEI test results, and the improved techniques it has brought to ontology mapping, DSSim can be considered as a state-of-the-art ontology mapping solution that compares well to other approaches.

However, this work also identifies limitations and further opportunities for research, which are also discussed in detail.

1. http://www.oracle.com/products/middleware/odi/docs/data-integration-market-whitepaper.pdf accessed 1st June 2010

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions