Exploring the Human-AI Continuum in Terminology Management: Towards Evaluation Criteria for LLM-Assisted Terminology Work

This paper explores two terminology management activities that lie at opposite ends of the automation spectrum within the terminology workflow: term extraction, a largely automatable, repetitive task in which human oversight can be limited, and definition generation, a conceptually demanding activity that requires strong human involvement. Based on initial considerations on the effective and appropriate use of LLM-generated outputs, the paper investigates the key aspects to consider when applying LLMs to both tasks. The outcome of this analysis is a set of evaluation criteria for LLM-assisted term extraction and LLM-assisted definition generation, with particular emphasis on positioning these tasks along the human -AI continuum in order to support quality, reliability, and accountability in terminology work.