Designing Metadata with Existing Ontologies

Semantic Web is expected to become an enhanced information foundation for both humans and machines. Once data on the WWW is annotated with machine understandable metadata linked to ontology, complex tasks can be automated using software agents. Among the issues surrounding the Semantic Web, difficulties in designing metadata act as a bottleneck to widespread adoption of the Semantic Web. This is due to the restrictions inherent in the metadata design task: Both humans and machines can create metadata, but in the case of the former, domain knowledge and design skill is required. In the case of the latter, metadata with only limited expressivity can be created. To tackle the problem of designing metadata, many existing researches propose design methodologies that focus on designing new metadata from scratch. Although most of them agree on the importance of reusing exiting ontologies, few give detailed solutions. This research takes up this less-tackled problem of reusing existing ontologies. The goal is to propose a metadata design method that utilizes existing ontologies as design material, and to specify what information, if provided in each design step, will be helpful to metadata designer in making design decisions. To achieve this, 1) existing ontologies on the WWW should be analyzed. This is necessary because these ontologies later constitute the design material. Unless ontologies on the WWW are abundant in variety and in quantity, referring to them would not be useful. Once the existence of various ontologies is confirmed, 2) problems in metadata design process need to be clarified. In particular, information needs that arise in a restricted setting, which prohibits human designer from referring to existing ontologies, should be investigated to highlight what information support may be useful. Based on the information needs obtained, 3) a step by step metadata design process which incorporates existing ontologies should be modeled, and design material metadata designer could refer to should be proposed. Firstly, to see what ontologies actually exist on the WWW, ten ontologies containing Person Class were analyzed with a focus on person related class definitions, property definitions and class hierarchies. Secondly, to clarify what problems arise in the design process, metadata design experiment was conducted by the author. As data, a social networking service's user input form was selected. A restriction was placed on the author by refraining from referring to existing ontologies. This was done to highlight information needs that arise during the design process. Lastly, …