[Paper Review] What Generative Artificial Intelligence Means for Terminological Definitions

Antonio San Martín|arXiv (Cornell University)|Feb 25, 2024

linguistics and terminology studiesArts and Humanities3 citations

TL;DR

This paper investigates how Generative Artificial Intelligence (GenAI), particularly ChatGPT, transforms the creation and use of terminological definitions by enabling context-sensitive, interactive, and customizable definitions that surpass traditional static definitions in usability. It proposes AI-assisted terminography—blending AI efficiency with human expertise—to accelerate definition generation, enhance flexibility, and improve quality, while acknowledging limitations in accuracy and reliability that preserve the enduring value of human-curated terminological resources.

ABSTRACT

This paper examines the impact of Generative Artificial Intelligence (GenAI) tools like ChatGPT on the creation and consumption of terminological definitions. From the terminologist's point of view, the strategic use of GenAI tools can streamline the process of crafting definitions, reducing both time and effort, while potentially enhancing quality. GenAI tools enable AI-assisted terminography, notably post-editing terminography, where the machine produces a definition that the terminologist then corrects or refines. However, the potential of GenAI tools to fulfill all the terminological needs of a user, including term definitions, challenges the very existence of terminological definitions and resources as we know them. Unlike terminological definitions, GenAI tools can describe the knowledge activated by a term in a specific context. However, a main drawback of these tools is that their output can contain errors. For this reason, users requiring reliability will likely still resort to terminological resources for definitions. Nevertheless, with the inevitable integration of AI into terminology work, the distinction between human-created and AI-created content will become increasingly blurred.

Motivation & Objective

• To examine the impact of Generative AI on the creation and consumption of terminological definitions, especially in comparison to traditional terminological resources.
• To explore how GenAI tools like ChatGPT can support terminographers in producing context-sensitive, flexible definitions more efficiently.
• To assess the potential of AI-assisted terminography as a new paradigm that integrates AI efficiency with human expertise in terminology work.
• To identify the limitations of GenAI in accuracy and reliability, and to understand why traditional terminological resources remain essential for high-stakes applications.
• To propose a future where AI-assisted terminography becomes standard practice, redefining the role of terminologists in the age of large language models.

Proposed method

• Using a cognitive linguistics framework, the paper conceptualizes terms as access points to semantic potential, with premeanings as contextually constrained subsets of this potential.
• The study evaluates ChatGPT’s ability to generate context-specific definitions by prompting it with specific texts, images, or described situations to elicit meaning-relevant responses.
• It proposes using corpus tools like Sketch Engine to extract concordance lines and word sketches (including semantic sketches) as structured input for prompting GenAI.
• The method involves post-editing AI-generated definitions using corpus-derived evidence, ensuring traceability and justifiability of content.
• Terminologists are advised to use ChatGPT for iterative refinement—evaluating, enhancing, and validating definitions through AI feedback.
• A novel evaluation method is proposed: using ChatGPT’s ability to guess the term from a definition to test its adequacy, with failure signaling the need for revision.

Experimental results

Research questions

• RQ1 To what extent can GenAI tools like ChatGPT generate context-specific, user-tailored terminological definitions that surpass traditional static definitions in usability?
• RQ2 How can AI-assisted terminography be implemented to combine the efficiency of GenAI with the reliability of human expertise in terminology work?
• RQ3 What are the limitations of GenAI in accuracy and consistency, and how do they affect the long-term viability of AI-generated definitions in specialized domains?
• RQ4 In what ways can GenAI tools enhance the definition-writing process, including content validation, error detection, and stylistic improvement?
• RQ5 How might the integration of AI into corpus analysis tools streamline the generation of flexible, evidence-based terminological definitions?

Key findings

• ChatGPT can generate context-specific definitions that reflect the meaning activated in a particular usage event, going beyond the premeaning described in traditional definitions.
• GenAI enables interactive, user-adaptive definition generation, allowing follow-up questions, clarifications, and contextual customization that static resources cannot provide.
• Despite its strengths, GenAI faces reliability issues, including factual inaccuracies and hallucinations, which limits its use in high-stakes or precision-critical applications.
• AI-assisted terminography—using GenAI to post-edit and refine human-generated definitions—can significantly increase productivity and improve definition quality, especially for flexible, context-sensitive definitions.
• The integration of GenAI with corpus tools like Sketch Engine allows for structured, evidence-based definition generation using concordance lines and semantic sketches, enhancing traceability and justification.
• The ability of ChatGPT to reverse-identify a term from a definition offers a novel, automated method for evaluating definition adequacy, with failure indicating the need for revision.