Abstract |
In this paper, we report on our development of a multi-level analysis framework that allows us to assess AI-generated lexicographic texts on both a quantitative and qualitative level and compare them with human-written texts. We approach this problem through a systematic and fine-grained evaluation, using dictionary articles created by human subjects with the help of ChatGPT as an example. The levels of our framework concern the assessment of individual entries, a comparison with existing dictionary entries written by experts, an analysis of the writing experiment, and the discussion of AI-specific aspects. For the first level, we propose an elaborate evaluation grid that enables a fine-grained comparison of dictionary entries. While this grid has been developed for a specific writing experiment, it can be adapted by metalexicographical experts for the evaluation of all kinds of dictionary entries and all kinds of dictionary information categories. |
BibTex |
@inproceedings{euralex_2024_paper_25, address = {Cavtat}, title = {Multi-Level Analysis as a Systematic Approach to Evaluating the Quality of AI-Generated Dictionary Entries},isbn = {978-953-7967-77-2}, shorttitle = {Euralex 2024}, url = {}, language = {eng}, booktitle = {Lexicography and Semantics. Proceedings of the XXI EURALEX International Congress}, publisher = {Institut za hrvatski jezik}, author = {Evert, Stephanie and Ganslmayer, Christine and Rink, Christian}, editor = {Despot, Kristina Š. and Ostroški Anić, Ana and Brač, Ivana}, year = {2024}, pages = {317-335} } |