Spm: Preparing hierarchical data structures for schema matching

Schema pre-matching approaches prepare different schema elements for the matching step. Ideally words are extracted from schema elements' labels and the semantic correspondences are generated. Identifying the sense of words based on their contexts and before schema matching is critical for increasing the total amount of true matches and decreasing significantly the total amount of both missed matches and false matches. However this problem is very challenging since we often do not have complete and precise information about the meaning behind each element. In this paper we present SPM a Sense-based Pre-Matching approach. SPM has two key modules. First the Sets of Words Generator module generates from every schema element a set of words that describes its meaning. Second the Sense Identifier module selects the most appropriate meaning of words based on their contexts. Experimental results on real-world datasets show high effectiveness achieved by SPM. © 2020 ACM.