29 September 2026 to 3 October 2026
OeAW Main Seat
Europe/Vienna timezone

Effectiveness of AI for the Lexicography of Low-Resource Languages: The Dictionary of Georgian Neologisms

2 Oct 2026, 09:40
20m
01 | Johannessaal (01 | Johannessaal, OeAW Main Seat, 1st floor)

01 | Johannessaal

01 | Johannessaal, OeAW Main Seat, 1st floor

Austrian Academy of Sciences Dr. Ignaz Seipel-Platz 2 1010 Vienna

Speakers

Tinatin Margalitadze (Ilia State University) Giorgi Meladze (Ilia State University) Ketevan Mchedlishvili (Ilia State University) Tamar Laluashvili (Ilia State University) Giorgi Okropiridze (Ilia State University)

Description

The purpose of this paper is to present the results of a study investigating the effectiveness of AI in the compilation of the Dictionary of Georgian Neologisms, an ongoing project at the Centre for Lexicography and Language Technologies at Ilia State University. The data for the study were selected from a dataset of 1,700 neologisms identified in previous research that developed a semi-automatic method for detecting neologisms in Modern Georgian (Laluashvili & Margalitadze, 2025). Three major categories of Georgian neologisms were selected: (1) borrowings, including direct loans, morphologically adapted loanwords, and loan translations; (2) word formation, with a special focus on internally created Georgian formations; and (3) semantic evolution. The experiments conducted in this study employed ChatGPT 5.2, Gemini 3.0 Thinking, and Grok 4.1 Thinking in combination to mitigate model-specific bias in handling Georgian linguistic material. These models were applied to the primary task of evaluating the effectiveness of LLMs in identifying the meanings of neologisms and generating lexicographic definitions. The findings demonstrate that LLMs can serve as powerful lexicographic tools capable of effectively processing Georgian linguistic data. However, their performance and accuracy are highly dependent on the prompts designed and employed by lexicographers.

Presentation materials

There are no materials yet.