講座內(nèi)容：Biomedical ontologies andterminologies play a vital role in healthcare information management, dataintegration, and decision support. Quality issues in biomedical ontologies, if not addressed, can affect thequality of all downstream information systems relying on them as a knowledgesource. Thus Ontology quality assurance (OQA) is an indispensable part of theontology engineering cycle. However, it is labor-intensive and time-consumingto discover errors or inconsistencies by manual review of large biomedicalontologies. Effective, automated approaches for improving the quality ofbiomedical ontologies are needed to overcome the limitations of manual work. Inthis talk, I will present a non-lattice-based approach to detecting and miningpotential errors in SNOMED CT, the most comprehensive clinical health careterminology worldwide. I will introduce a scalable MapReduce pipeline forexhaustively extracting non-lattice pairs from SNOMED CT, and an effective methodfor mining lexical patterns in non-lattice subgraphs to detect errors in SNOMEDCT and suggest remediations. Since virtually all biomedical ontologies areorganized into subsumption hierarchies and have concept names, ournon-lattice–based approach can be generalized and applied to other biomedicalontologies for quality assurance purposes.