ChemDataExtractor 2.0: Autopopulated Ontologies for Materials Science

Area of Science:

• Chemistry
• Materials Science
• Data Science

Background:

• Increasing data in scientific publications necessitates automated extraction techniques.
• Shift from extracting individual properties to higher-level relationships in physical sciences.
• Need for methods to integrate primary literature into data-driven scientific frameworks.

Purpose of the Study:

• To present a framework for automated ontology population through direct extraction of property networks.
• To develop a model for extracting chemical and physical properties, including hierarchical and nested data.
• To demonstrate the framework's capability in extracting complex crystallographic information.

Main Methods:

• Exploiting data-rich sources like tables within scientific documents.
• Developing a novel model for hierarchical data organization and extraction.
• Utilizing automatically generated parsers and interdependency resolution.
• Applying the framework to extract crystallographic hierarchies from scientific articles.

Main Results:

• Successfully extracted 18 interrelated submodels of nested crystallographic data.
• Achieved an overall precision of 92.2% across 26 different scientific journals.
• Demonstrated the framework's effectiveness in handling complex, linked data.

Conclusions:

• The developed framework and toolkit (ChemDataExtractor 2.0) enable automated population of ontologies.
• This approach facilitates the seamless integration of primary scientific literature into data-driven systems.
• Offers a significant advancement in extracting and organizing scientific data for broader use.