Automated Information Extraction from Safety and Material Data Sheets—A Domain-Specific NLP Pipeline for Structured Material Data Management in Battery Cell Production

The performance of lithium-ion batteries is strongly determined by material properties, which are provided in technical data sheets but often in inconsistent formats and terminology. Automated extraction of these parameters could enable downstream applications such as process optimization, traceability, and hazard assessment. However, current approaches are unsuitable for industrial use. This work presents a prototype NLP-based extraction pipeline for material and safety data sheets. Using fine-tuned SpaCy models, F1-scores above 0.7 are achieved for key parameters such as CAS number, molecular mass, and density. The resulting structured material database provides a foundation for data-driven applications in battery cell production. The feasibility of domain-specific NLP for automated material information extraction is demonstrated and potential pathways for integration with process control and optimization workflows are discussed.