Interoperating ontologies of organizational memory through hybrid unsupervised data mining

摘要

Purpose The purpose of this paper is to present an automated ontology mapping and mergingalgorithm, namely OntoDNA, which employs data mining techniques (FCA, SOM, K-means) to resolveontological heterogeneities among distributed data sources in organizational memory andsubsequently generate a merged ontology to facilitate resource retrieval from distributed resourcesfor organizational decision making. Design/methodology/approach – The OntoDNA employs unsupervised data mining techniques(FCA, SOM, K-means) to resolve ontological heterogeneities to integrate distributed data sources inorganizational memory. Unsupervised methods are needed as an alternative in the absence of priorknowledge for managing this knowledge. Given two ontologies that are to be merged as the input, theontologies' conceptual pattern is discovered using FCA. Then, string normalizations are applied totransform their attributes in the formal context prior to lexical similarity mapping. Mapping rules areapplied to reconcile the attributes. Subsequently, SOM and K-means are applied for semanticsimilarity mapping based on the conceptual pattern discovered in the formal context to reduce theproblem size of the SOM clusters as validated by the Davies-Bouldin index. The mapping rules arethen applied to discover semantic similarity between ontological concepts in the clusters and theontological concepts of the target ontology are updated to the source ontology based on the mergingrules. Merged ontology in a concept lattice is formed. Findings – In experimental comparisons between PROMPT and OntoDNA ontology mapping andmerging tool based on precision, recall and f-measure, average mapping results for OntoDNA is 95.97percent compared to PROMPT's 67.24 percent In tetras of recall, OntoDNA outperforms PROMPT on allthe paired ontology except for one paired ontology. For the merging of one paired ontology, PROMPTfails to identify the mapping elements. OntoDNA significantly outperforms PROMPT due to theutilization of FCA in the OntoDNA to capture attributes and the inherent structural relationships amongconcepts. Better performance in OntoDNA is due to the following reasons. First, semantic problems suchas synonymy and polysemy are resolved prior to contextual clustering. Second, unsupervised data miningtechniques (SOM and K-means) have reduced" problem size. Third, string matching performs better thanPROMPT's linguistic-similarity matching in addressing semantic heterogeneity, in context it alsocontributes to the OntoDNA results. String matching resolves concept names based on similarity betweenconcept names in each cluster for ontology mapping. Linguistic-similarity matching resolves conceptnames based on concept-representation structure and relations between concepts for ontology mapping. Originality/value – The OntoDNA automates ontology mapping and merging without the need ofany prior knowledge to generate a merged ontology. String matching is shown to perform better thanlinguistic-similarity matching in resolving concept names. The OntoDNA will be valuable fororganizations interested in merging ontologies from distributed or different organizational memories.For example, an organization might want to merge their organization-specific ontologies withcommunity standard ontologies.