Named Entity Recognition and Normalization Applied to Large-Scale Information Extraction from the Materials Science...

Area of Science:

• Materials Science
• Computational Materials Science
• Data Science

Background:

• The rapid growth of materials science literature presents a challenge in synthesizing existing knowledge.
• Connecting new research findings with established literature is a bottleneck in materials discovery.

Purpose of the Study:

• To develop a text mining framework for large-scale information extraction from materials science literature.
• To create a structured database of materials science knowledge for programmatic querying.

Main Methods:

• Application of text mining with named entity recognition (NER) for information extraction.
• Training an NER model to identify inorganic materials, sample descriptors, properties, synthesis, and characterization methods.
• Processing 3.27 million materials science abstracts to extract over 80 million named entities.

Main Results:

• Achieved 87% accuracy (f1-score) in extracting summary-level information.
• Converted millions of abstracts into structured database entries.
• Demonstrated the ability to answer complex literature-based questions via simple database queries.

Conclusions:

• The developed system effectively extracts and structures information from materials science literature.
• This approach accelerates materials discovery by enabling efficient querying of vast amounts of data.
• The data and tools are publicly available to foster further research and development.