LD Connect: A Linked Data Portal for IOS Press Scientometrics

In this work, we describe a Linked Data portal, LD Connect, which operates on all bibliographic data produced by IOS Press over the past thirty-five years, including more than a hundred thousand papers, authors, affiliations, keywords, and so forth. However, LD Connect is more than just an RDF-based metadata set of bibliographic records. For example, all affiliations are georeferenced, and co-reference resolution has been performed on organizations and contributors including both authors and editors. The resulting knowledge graph serves as a public dataset, web portal, and query endpoint, and it acts as a data backbone for IOS Press and various bibliographic analytics. In addition to the metadata, LD Connect is also the first portal of its kind that publicly shares document embeddings computed from the full text of all papers and knowledge graph embeddings based on the graph structure, thereby enabling semantic search and automated IOS Press scientometrics. These scientometrics run directly on top of the graph and combine it with the learned embeddings to automatically generate data visualizations, such as author and paper similarity over all journals. By making the involved ontologies, embeddings, and scientometrics all publicly available, we aim to share LD Connect services with not only the Semantic Web community but also the broader public to facilitate research and applications based on this large-scale academic knowledge graph. Particularly, the presented scientometric system generalizes beyond IOS Press data and can be deployed on top of other bibliographic datasets as well.