Migrating bibliographic datasets to the Semantic Web: The AGRIS case

AGRIS is among the most comprehensive online collections of agricultural and related sciences information. It is a growing global catalog of 5 million high-quality structured bibliographic records indexed from a worldwide group of providers. AGRIS relies heavily on the AGROVOC thesaurus for its indexing. Following the conversion of that thesaurus into a SKOS concept-scheme and its publication as Linked Open Data (LOD), the entire set of AGRIS records was also triplified and released as LOD. As part of this exercise, OpenAGRIS, a semantic mashup application, was developed to dynamically combine AGRIS data with external data sources, using a mixture of SPARQL queries and web services. The re-engineering of AGRIS for the Semantic Web raised numerous issues regarding the relative lack of administrative metadata required to compellingly address the proof and trust layers of the Semantic Web stack, both within the AGRIS repository and in the external data pulled into OpenAGRIS. The AGRIS team began a process of disambiguation and enrichment to continue moving toward an entity-based view of its resources, beginning with the tens of thousands of journals attached to its records. The evolution of the system, the issues raised during the triplification process and the steps necessary for publishing the result as LOD content are hereby discussed and evaluated.