Extracting Chemical Information from Scientific Literature Using Text Mining: Building an Ionic Conductivity Database...

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Growing demand for electric vehicles (EVs) increases safety concerns with traditional liquid electrolytes.
• Liquid electrolytes in rechargeable batteries pose fire and explosion risks.
• Solid-state electrolytes (SSEs) offer greater stability, driving research for high ionic conductivity materials.

Purpose of the Study:

• To automate the extraction of ionic conductivity data for SSEs from scientific literature.
• To construct a comprehensive materials database for accelerating SSE discovery.
• To overcome the time-consuming and repetitive nature of manual data collection.

Main Methods:

• Developed a text-mining pipeline including document processing, natural language preprocessing, phase parsing, and relation extraction.
• Applied the model to extract ionic conductivities from 38 research studies for performance verification.
• Successfully distinguished ionic from electrical conductivities, reducing undistinguished records from 93% to 24.3%.

Main Results:

• Validated the accuracy of the text-mining model by comparing extracted conductivities with actual values.
• Significantly improved the ability to differentiate ionic conductivity from electrical conductivity in battery research data.
• Constructed a large-scale ionic conductivity database from 3258 papers.

Conclusions:

• Automated text mining is an effective method for rapidly building materials databases for SSEs.
• The developed database accelerates the exploration and discovery of new, stable SSEs for safer EV batteries.
• This approach enhances data accessibility and research efficiency in solid-state battery development.