Automated trait extraction from unstructured species descriptions for species ID: a pilot study using a large language model

Digitised collections of floral descriptions are a major source of trait information for digital identification systems. Mobilising species descriptions requires manual trait extraction and conversion to a structured format, which can significantly increase time and resource demands. We explored if a large language model (LLM) trained to process unstructured textual data could be used to automate and therefore accelerate the trait extraction pipeline for species identification.