Semantic Schema Matching for String Attribute with Word Vectors and its Evaluation

Abstract

Instance-based schema matching is to determine the correspondences between heterogeneous databases by comparing instances. Heterogeneous databases consist of an enormous number of tables containing various attributes, causing the data heterogeneity. In such cases, it is effective to consider semantic information. In this paper, we propose the instance-based schema matching considering attributes’ semantics. We used Word2Vec to match attributes of character strings. The result shows a possibility to detect matching between attributes with high semantic similarity.