Interactive Refinement of Linked Data: Toward a Crowdsourcing Approach

摘要

This paper proposes an approach in which a system extracts information interactively from a user to refine linked data. A multilingual frequently-asked-questions (FAQs) database in the domain of rental apartments is used as a test-bed. This database includes a domain ontology represented using linked data. It contains the relationships between part of the floor plan of a rental apartment and FAQ entries, which are derived using the domain ontology. When a user finds an error in the relationship, an interactive process is initiated so that even a casual user who is not a domain expert can contribute to fixing the error. The proposed method also makes use of pictures so that the language barrier can be lowered. Since the proposed approach is targeted to a casual user, it can be used in crowdsourcing the refinement of linked data.