Estimating Domain Models from Metadata Instances to Improve Usability of LOD Datasets

Ryouta Kinjou, Mitsuharu Nagamori, Shigeo Sugimoto

Abstract


Linked Open Data(LOD), which is one of the efforts to help realize semantic web, has gradually become popular. Many Linked Open Data datasets, however are not well utilized. There are multiple reasons for this, such as a low level of recognition of LOD, limited usability of LOD datasets and so on. In attempting to solve these issues, we focused on a metadata schema that describes the structure about metadata instances in each LOD dataset. As information about metadata schema are not typically released, it is difficult to use LOD datasets. Therefore, in this research we extract the domain model, which is one piece of information about a metadata schema, from metadata instances. Domain models are suitable for understanding the rough structure of a metadata instances in an early stage. We developed an estimation method to generalize a process of understanding metadata schema when people, who are not familiar to the datasets, deal with. We then apply the estimation method to existed datasets.