統合モビリティサービスの実現に向けたスキーママッチング手法の提案

抄録

We propose a schema matching method of transport service data in different schema from multiple data sources. Nowadays, open data about transport services are increasing. However, there are some challenges to use them, because they have complicated data structure such as deeply nested objects or linked small objects. And their disparate data structures make data mathcing more difficult. So we present a method to make match trasnport services data by exploit the data strucutre, terminology and domain knowledge. The proposed method gradually add linked objects to matching target in mathcing process to find corresponding attributes. Generally, wider matching target space raises false positive rate. So our proposed method caluculates similarity score of attributes taking into account reference relationships between linked objects. We experimentaly evaluate the method in real-world transport service data and comfirm the method can improve mathcing accuracy.