Semantic Structure Matching Recommendation Algorithm

The paper presents a new hybrid schema matching algorithm: Semantic Structure Matching Recommendation Algorithm (SSRMA). SSRMA is able to discover lexical correspondences without breaking the structural ones — it is capable of rejecting trivial lexical similarities, if the structural context suggests that a given matching is inadequate. The algorithm enables achieving results that are comparable to those obtained by means of state-of-the-art schema matching solutions. The presented method involves an adaptable pre-processing and flexible internal data representation, which allows to use a variety of auxiliary data (e.g., textual corpora) and to increase the accuracy of semantic matches accommodated in a given domain. In order to increase the mapping quality, the method allows to extend the input data by auxiliary information that may have the form of ontologies or textual corpora.