Exploring the use of text mining to identify firms' green business process patterns

The vast amount of unstructured textual data on firms’ sustainability efforts poses major analytical challenges for sustainability researchers. Text mining techniques offer promising solutions to overcome the limitations of manual coding, enabling analysis on a larger scale. This study compares keyword-based searches against Large Language Models (LLMs) applications to extract and map sustainable business process patterns from firms’ sustainability reports. Results show that integrating LLMs with manual coding improves precision and accuracy over automated or keyword-based approaches alone. However, challenges persist, including variability in LLM outputs, extraction errors, and hallucinations, underscoring the limitations of relying solely on automated methods.