Uses WildChat
“An bhfuil Gaeilge agat?”: Differences in User Interaction and Assistant Responses Across Languages of European Origin in Large-scale Conversational Datasets
AICS 2024 · Proceedings published
Compares topics, conversation length, sentiment, moderation scores, and inferred satisfaction across 28 languages using WildChat and LMSYS-Chat-1M. The study combines corpus analysis with a classifier for user–assistant–user interaction triples.
Data used
WildChat (990,372 conversations) combined with LMSYS-Chat-1M; language and quality filters leave 781,376 conversations across 28 languages of European origin.
Findings
Topics substantially overlap across language groups, while conversation length and moderation scores vary. A model estimates satisfaction from user–assistant–user interaction triples.
Section 2: dataset combination and filtering. Sections 3–7: topic, length, sentiment, moderation, and satisfaction analyses. Read the source.
Dates and versions
CEUR proceedings publication date; presented at AICS 2024 on December 9–10, 2024. Proceedings record.
The conference took place in December 2024; CEUR published volume 3910 in March 2025.
Cite this paper
Aldan Creo. 2025. “An bhfuil Gaeilge agat?”: Differences in User Interaction and Assistant Responses Across Languages of European Origin in Large-scale Conversational Datasets. Proceedings of the 32nd Irish Conference on Artificial Intelligence and Cognitive Science.
Show BibTeX
@inproceedings{creo2025gaeilge,
title = {{“An bhfuil Gaeilge agat?”: Differences in User Interaction and Assistant Responses Across Languages of European Origin in Large-scale Conversational Datasets}},
author = {Creo, Aldan},
year = {2025},
booktitle = {Proceedings of the 32nd Irish Conference on Artificial Intelligence and Cognitive Science},
publisher = {CEUR Workshop Proceedings},
volume = {3910},
pages = {202--214},
url = {https://ceur-ws.org/Vol-3910/aics2024_p83.pdf}
}