For many years the majority of digitally recorded data, both scientific and the like, has been encoded in formats that are aligned for storage in relational database management systems such as MySQL, SQLite, and MS SQL. Under such systems, data is organized in a tabulated fashion where a table or relation may be created for a chemical compound and its columns and rows, or records, may correspond to the compound’s respective properties and values. The SciData Project aims to establish a new standard for the semantic representation scientific data using Resource Description Framework (RDF) syntax within an open-sourced Graph Database framework powered by Apache Jena, made accessible through the Jena Fuseki SPARQL Endpoint. As opposed to storing information categorically in tables, data is ingested into a graph consisting of interconnected networks of nodes and edges allowing for unprecedented levels of query complexity. Furthermore, by defining basic concepts and relationships through standardized ontologies, inferencing tools may then be applied to an established database to discover new connections between previously uncorrelated nodes of information- allowing for data-driven discovery that is simply not possible in traditional database models. The key to the SciData ecosystem begins with a JSON-LD (JavaScript Object Notation for Linked Data) file. Much like the widely adopted XML format, JSON-LD is a highly compatible and lightweight format that is set aside by its flexibility to carry scientific data ranging from numerical property data of a substance to the author and citation information of the data itself. Information, of varied sources and formats, must first pass through SciFlow- a custom python project that has been written to first convert, then validate the source data as an acceptable JSON-LD document. Once the data has been validated to meet pre-specified quality standards, it is then injected into a database via W3C standard HTTP-POST requests where it is parsed and integrated into a graph to be inferenced and queried. This talk will focus on deploying Apache Jena and ingesting JSON-LD files in order to be able to perform chemically relevant semantic searches across initially heterogeneous data.

Semantic searching of SciData JSON-LD format chemical property data

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.