Description
This article presents an automated approach to integrate multiple analogous ontologies extracted from structured web pages into a common ontology. These ontologies from heterogeneous systems exhibit rich diversity in appearances, structures, terminologies and granularities. We design a unified similarity paradigm that can collect the implicit and explicit evidences that exhibit coherences among ontology and instance, semantic and structure, as well as linguistic and syntactic features. The similarity between ontology elements is derived from three aspects such as intension, extension and context, denoted by , where INT and EXT include corresponding weighted contents from their offspring, and CXT is relevant to evidences shown in their ancestors. The similarity in each aspect is calculated by means of their semantic overlapping and syntactic comparability. We develop a top-down matching algorithm based on matching space selection and similarity reuse; the algorithm facilitates less error-prone map-pings and lower computational cost.