Semantic Based Database Schema Matching

matching is a process that takes multiple schemas as an input, applying a mapping between these schemas and produces one schema with the matching components. Most of data sources of E-business companies are heterogeneous, which making it difficult task for the integration and the exchange of the data. The purpose of this thesis is to indicate a matching system that involves required steps to produce a map while matching relational databases to ontology. According to researchers schema matching using two main approaches for classification; Individual Matcher Approaches and Combining Matcher Approaches with multiple levels of matching criteria. In this thesis individual matcher based on schema with linguistic based is used. A research done to find the suitable methodology that works effectively with schema matching; Conventional Neural Network (CNN) indicated to extract the important features and classify it to get a match. CNN uses word-embedding representations of two words as inputs to derive the semantic characteristics between the two words and provide a score as the result of how likely they fit the CNN pattern, Cosine Similarity and Jaro Winkler algorithms implemented with CNN to get better results. The combination of semantic schema matching with machine learning algorithm provide great improvement in the matching field, which become easier and time consuming. The performance of the proposed system showed that semantic similarity scores with clustering using CNN model could produce more than 90% accuracy. NetBeans and protégé are used to build the proposed system. Keywords: Semantic Matching, Ontology, Machine Learning, CNN, Word Embedding, Backpropagation.