SpringerMaterials Accelerates Research With Semantic Search

The Challenge

To provide a seamless experience across all the data sources within SpringerMaterials, the Database Group manages the tools and infrastructure that serve as the backbone to the platform. To bring together the variety of discrete data sources that are incorporated into SpringerMaterials, the DG team originally relied on robust XML document store and keyword/element builds. However, the legacy infrastructure required time-intensive updates and was outgrown by the demands of users and internal needs. As the team brought on more data sources to integrate into the platform, developed novel features for search and interaction, and updated existing data with new research, the connections within the infrastructure strained under the pressure of the ever-evolving data slowing down SpringerMaterials release cadence.

It became obvious that the semantic search functionality subscribers relied on could no longer be supported by the existing data structure. Although there was robust data underlying each search, there was potential to develop infrastructure that permits users to see the dynamic connections within and relationships between different materials, properties, and publications. The limitations to semantic search also prevented organic exploration within the data and obscured the data links that provide context to the researchers. To meet the internal maintenance and development needs of the Database Group, and allow external users to maximize the value from their immense pool of data within SpringerMaterials, the team needed to refine and replace their traditional XML document store backend.

The Solution

To meet end-user and internal requirements, the Database Group needed a flexible and dynamic data infrastructure that allowed easy iterations and updates within existing data while also providing context embedded within search results to end-users. By unifying their data in a knowledge graph, the Database Group created a single declarative model to bring all the connections between the content into a unified view.

Going live 18 months after inception, the SpringerMaterials knowledge graph contains 100 million triples spanning data sources like the Landolt-Börnstein series, Springer Handbooks, and relational databases that cover topics from corrosion to polymers. With Stardog’s native flexibility and scalability, the Database Group team updates the knowledge graph on quarterly intervals and rolls out functionality enhancements without the constraints the previous infrastructure put on the team. Further, sub-second query times get users the information they need seemingly instantaneously.

The flexibility and dynamic connections inherent within the SpringerMaterials knowledge graph now power a semantic search capability that understands user’s intent and can uncover complex relationships from simple keywords. Built alongside the knowledge graph, these new intelligent search features now include:

• On-the-fly interpretation of search context and users’ intent
• Adaptive ranking of results based on search context
• Entity-based search and explore capabilities
• Reverse search to find substances based on property values
• Smart light-weight inference on substance and property identifiers
• Plug and play integration of external knowledge models
• Dynamically created landing pages

The Results

With 50,000 users, SpringerMaterials can facilitate and enhance materials research by providing their subscribers on-demand and interconnected information. Users can now:

• Access one of the most comprehensive, multidisciplinary collections of materials property data—more than 290,000 materials—across the research sciences
• Expedite research with time-saving queries that provide accurate and focused search results
• Easily integrate materials data into various research workflows

Going forward, the Database Group will expand the connections within the data beyond the data sources within SpringerMaterials to the data sources that span the rest of the Database and Solutions offerings. Two other Springer Nature databases, Springer Nature Experiments and Nano — a Nature Research solution, are also powered by Stardog. In the future, users will be able to federate their searches across all of their database subscriptions. The native flexibility of a knowledge graph allows Springer Nature to make continuous additions and adjustments to the data as more discoveries are made, their proprietary datasets grow, and new data sources emerge.