Schema Matching Using Word-level Clustering for Integrating Universities’ Courses

Schema matching is the process of determining the similarities among multiple schemas of databases or websites structure. Several matching approaches have been depicted in the literature. The most adequate and accurate approach is related to the textual analysis. In this approche, the schemas are being examined in terms of its text information. Various studies used textual-based matching by addressing the character-level string similarity. However, the character-level similarity would have some limitations regarding the words with similar characters but different meanings. Therefore, this study aims to propose a word-level similarity method based on term frequency and cosine similarity. Consequentially, a K-means clustering will utilize the proposed matching analysis to group the correspondences among the schemas. A dataset for universities’ courses is being used to test the proposed method. Experimental results showed that the proposed word-level similarity method had outperformed the character-level by obtaining 73% accuracy. This indicates the significance of using word-level similarity when using cluster-based schema matching.