Looking through glass: Knowledge discovery from materials science literature using natural language processing.

Area of Science:

• Materials Science
• Computational Materials Science
• Data Science

Background:

• The vast majority of materials science knowledge is locked in unstructured text and image formats within scientific literature.
• Efficiently extracting and organizing this information is crucial for accelerating research and discovery.

Purpose of the Study:

• To develop an automated framework for comprehending and extracting precise knowledge from inorganic glasses literature.
• To enhance the accessibility and searchability of scientific publications and their associated data.

Main Methods:

• Utilizing natural language processing (NLP) for text and image analysis.
• Employing Latent Dirichlet Allocation (LDA) for automatic abstract categorization and semantic linking of publications.
• Implementing Caption Cluster Plot (CCP) for summarizing and accessing images and plots within papers.

Main Results:

• Successful automated categorization of research abstracts based on semantic content.
• Generation of comprehensive image summaries, enabling direct access to visual data.
• Creation of an 'elemental map' by integrating LDA, CCP, and chemical elements, showing topical and image-wise distribution of elements.

Conclusions:

• The presented framework offers a generic and powerful tool for extracting and disseminating materials-specific information.
• This approach facilitates insights into composition-structure-processing-property relationships.
• The framework has the potential to accelerate the discovery of novel materials.