Research On Taxonomic Relationships Evaluation In Ontologies Based On Semantic Similarity

As a concept model which can describe information and data on the semantic level, ontologies are already widely used in data annotations, data integration, and data intelligence in recent years. However, most of these ontologies have been developed from scratch by individuals or institutions of specialists, which is the reason that it is difficult for ontologies to interoperate. As early as 1990s, the need for an evaluation methodology in the field of ontology development and reuse has become pressing and since then the need has always been growing accompanied by the development trend of information technology. The works of ontology evaluation can be carried out from two fields: such as, the study to evaluate the lexical layer in ontologies, existing quite a few typical methods; the research to evaluate the taxonomy in ontologies, using formalistic ways. But in general, the existing methods not only lack of a complete system to evaluate ontology taxonomy, but also ignore the semantic level of the evaluation.In this paper, we propose a new measure method for evaluating taxonomy in ontologies using lexical semantic similarity. According to the relativities of concepts, we describe the way to estimate the consistency and redundancy of ontology taxonomy.The main works in this paper are detailed as follow:1. Firstly, we make a WordNet-based similarity computing algorithm - DWN. The algorithm can be used to calculate the semantic similarity between the concept pair in the ontology .2. Secondly, we propose a new method for evaluating ontologies taxonomy based on semantic similarity - Con2Onto, which resolves the inconsistency errors and redundancy errors happened in the taxonomy of the ontology. In the process of consistency evaluation, it finds not only the cyclization errors and semantic errors, but also the division errors of the parent concept. In the evaluate process for redundancy, two different algorithms have designed to evaluate the concept redundancy and the syntactic redundancy.3. Lastly, as a part of this research, we have applied our approach in evaluating the taxonomy in Expo domain ontology, and then based on the application results, we design a set of measure metrics to show that our approach is effective.