多维度相似度融合在OAEI的应用与评估OA

In view of the lack of systematic evaluation of various similarity fusion strategies in the existing ontology mapping methods,we propose a standardized evaluation framework,aiming to objectively compare the effectiveness of different fusion strategies in ontology matching.The research calculates similarity from four dimensions:semantics,strings,attributes,and instances,and integrates the similarity matrix by adopting seven fusion strategies including maximum value,average value,Sigmoid,Softmax,principal component analysis,and efficient adaptive weights.The matching generation uniformly adopts the stable marriage algorithm to ensure the stability and comparability of the pairing results.The experiment is based on the benchmark dataset of OAEI and is tested in four scenarios:complete structure,missing attributes,missing instances,and double missing,with the F1 value as the core evaluation index.The results show that only in the specific situation where the ontology lacks instance data,the Sigmoid function method exhibits relatively better adaptability.In most other scenarios,the efficient adaptive weighting method significantly outperforms other fusion strategies in terms of accuracy and ro-bustness,providing quantifiable empirical evidence for the selection and optimization of similarity fusion strategies in ontology mapping.