Automated Ontology Alignment
In view of the need of highly distributed and federated architecture, integrating schemas from different data sources in a specific domain has great impact on the performance of web applications. As the web scales, there are ample sources that provide structured information (ontologies) in the same domain. Since ontologies for a same domain usually overlap, we aim to determine a global ontology based on the commonality of overlapping entities using an efficient algorithm. Our algorithm examines ontologies by considering entities and relationships between them in ontologyâ€™s graph. First it finds out the Largest Common Subgraph (LCS) between two ontologies. For this, LCS is aligned to other ontologies using the lexical and structural similarity of entities. Next, we propose a novel statistical model based on maximum likelihood to determine a global ontology. This statistical model exploits the commonality of each entity across different ontologies. It will be helpful for federated query expansion.
Professors: Dr. Latifur Khan, Steven Seida and Dr. Bhavani Thuraisingham
Students: Neda Alipanah and Julie Rauer
Documents and Publications
- Technical Report (UTDCS-04-10)
- Technical Report (UTDCS-27-10)