LogMap: Logic-based Methods for Ontology Mapping

Ontologies are extensively used in biology and medicine. A prominent example of a bio-medical ontology is SNOMED CT, which is a core component of the NHS patient record service. Other examples include the Foundational Model of Anatomy (FMA) and the National Cancer Institute Thesaurus (NCI). Ontologies such as SNOMED CT, FMA, and NCI are gradually superseding the existing medical classifications and are becoming core platforms for accessing, gathering, and sharing medical knowledge and data.

To exchange or migrate data between ontology-based applications, it is crucial to establish correspondences (or mappings) between their ontologies. Creating such mappings manually is often unfeasible due to the size and complexity of modern ontologies. Therefore, the problem of automatically generating mappings between ontologies (often referred to as the ontology matching, ontology alignment, or ontology mapping problem) has been investigated extensively in recent years. Despite the already mature state of the art, bio-medical ontologies still pose serious challenges to existing techniques.

First, carefully-curated mapping sets used in bio-medical information integration, such as UMLS Metathesaurus, often contain errors and lack important information. Second, existing mapping generation tools do not scale to the size of modern bio-medical ontologies. Finally, the developers of bio-medical information integration and migration systems based on ontologies currently lack the necessary tool support.

In this research, we aim to address each of these needs. Therefore, we have identified three main objectives:

1. To develop general principles and specific techniques to efficiently detect and repair potential errors and discover missing information in manually-curated mapping sets, such as the UMLS Metathesaurus.
2. To develop computationally efficient algorithms for generating mappings between large-scale ontologies, while minimising the number of errors and the amount of missing information.
3. To design and implement a prototype with core functionality for mapping generation and management, as well as for the translation and migration of ontology data.

Our main research hypothesis is based on the observation that existing techniques for ontology mapping often disregard the logic-based semantics of the input ontologies. As a result, they fail to take advantage of the available semantics, and of the highly effective reasoning services for modern ontology languages. We are proposing to rethink the foundations underlying the current state-of-the art in the field by incorporating reasoning-based techniques in each of the steps of the ontology mapping process. We also intend to go even further and make our techniques practical and ready to be used in applications. To this end, we aim to develop suitable heuristics that can be efficiently implemented. The research is based on our preliminary empirical evidence which suggests the potential benefits of logic-based heuristic techniques when analysing existing mappings between biomedical ontologies.

We expect that our results will be directly relevant to the users of ontology-based systems in the bio-medical domain, where knowledge and data integration is a matter of major concern.