External knowledge sources are commonly used in processing large amounts of data. Large external knowledge sources, such as ontologies, often contain hundreds of thousands of concepts and relationships, making comprehension and navigation difficult. Abstraction networks enhance the usability and comprehensibility of these resources by providing a higher level of abstraction. In this paper, we develop a methodology to generate an ontology's abstraction network based on the semantic similarity measure of concepts. We first score the level of similarity between concepts based on shared hierarchical and semantic relationships. We then partition the original ontology into groups based on the concept pairs' similarity in relation to a user set threshold. We apply our methodology on the SNOMED CT Specimen hierarchy to generate three levels of abstraction.
Cirella D., Gu H. (2017, November). Generating abstraction networks using semantic similarity measure of ontology concepts. Paper presented at the 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Kansas City, Mo.; Proceedings of the 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 840-843. doi: 10.1109/BIBM.2017.8217764.