What Generative Artificial Intelligence Means for Terminological Definitions

Introduction

This paper discusses how terminological definitions are an essential part of specialized lexical resources, but their creation is time-consuming. The classical approach defines terms based on universal and context-independent characteristics, but this approach has drawbacks. The Flexible Terminological Definition Approach argues that accounting for context is necessary to create useful definitions.

From a cognitive linguistics perspective, terms derive meaning from context, not inherent properties. Terminologists must select the most relevant information from a term's vast semantic potential based on contextual and functional constraints to create definitions. This makes definition creation even more time-consuming.

Generative AI tools like ChatGPT can reduce the time and effort required to create definitions. However, AI could also have a more profound impact, potentially replacing terminological resources, terminologists, and terminology work tools. While it's too early to say for certain, the paper discusses factors that may influence the future of terminological definitions in the face of AI.

GenAI and the future of terminological definitions

The summary of the provided section of the paper is as follows:

The paper discusses the advantages and disadvantages of using ChatGPT, a generative AI, versus consulting a terminological resource to define terms. The main advantage of using ChatGPT is that it can describe the meaning of a term in a specific context, unlike a static definition. ChatGPT can tailor its responses based on the user's needs and provides an interactive experience to clarify and seek examples. Additionally, ChatGPT can offer more up-to-date information compared to terminological definitions. However, ChatGPT's responses may contain factual, logical, and linguistic errors, as well as biases and hallucinations. It also struggles with handling sense splitting and lumping. Furthermore, ChatGPT generates different responses to identical queries, posing challenges in situations requiring a single unchanging definition.

The paper suggests that the main advantage of a terminological definition is its reliability. Definitions with a prescriptive or standardizing orientation, as well as in-text definitions in publications, are less likely to be replaced by AI. While the terminologist profession may face challenges due to AI advancements, terminological resources created by humans are not likely to disappear, although their number may decrease due to commercial or funding considerations. The boundary between human-created terminological resources and AI-generated definitions is expected to become increasingly blurred.

AI-assisted terminography

The paper discusses the use of AI, specifically ChatGPT, in assisting terminologists with terminographic work, which is the compilation and management of terminology. This is referred to as "AI-assisted terminography." A particularly relevant form is "post-editing terminography," where the machine produces a terminographic item, such as a definition, which the terminologist then corrects or refines.

The paper outlines several benefits and challenges of using ChatGPT in post-editing terminography. On the one hand, it can enable terminologists to create definitions faster without compromising quality, allowing terminological resources to keep pace with the rapid emergence of new terms and concepts. However, the potential for errors and biases in ChatGPT's output requires careful verification, and the generated definitions may constitute plagiarism or reproduce copyrighted content.

To address these challenges, the paper discusses strategies for using ChatGPT more effectively, such as employing optimized prompts, providing contextual constraints, and instructing ChatGPT to follow definitional templates. It also notes that post-editing terminological definitions may currently only be viable in high-resource languages, as the quality of ChatGPT-generated definitions in less-resourced languages may not be adequate.

The paper also explores other forms of AI-assisted terminography, such as using ChatGPT to solve notional doubts, analyze existing terminological definitions, and assist with corpus analysis for definition writing. Additionally, ChatGPT can be used to refine terminologists' draft definitions by assessing and suggesting enhancements.

Overall, the paper suggests that the strategic use of ChatGPT and other AI technologies can enhance the efficiency and quality of terminological work, though it also highlights the need for careful oversight and validation to address the potential challenges.

Conclusions

The provided text discusses how AI has the potential to transform terminology work, which has historically been resistant to automation. AI-based tools, such as GenIA, are now able to generate terminological definitions that are often considered satisfactory by users, challenging the traditional role of terminological resources. This shift is expected to increase the efficiency of terminologists, allowing for faster updates to terminological resources and the creation of more flexible definitions that better reflect the contextual variability of terms. As AI technologies continue to advance, the quality gap between AI-generated and human-created content is likely to diminish, requiring terminologists to embrace AI and develop the necessary skills for its integration into their workflow. The text acknowledges that just as computers have become standard in terminography, AI-assisted terminography will likely become synonymous with the field.