2024/02/20 by Vincent Jung, Jung, Vincent, van der Plas, Lonneke · 1 citation
Arts and Humanities · Psychology · Social Sciences · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Linguistic Variation and Morphology #Linguistics, Language Diversity, and Identity #Stuttering Research and Treatment
paper · pdf · doi:10.48550/arxiv.2402.13016
openalex publication_date 2024/02/20 · openalex created_date 2024/02/22 · openalex updated_date 2026/07/28
We study the effect of one type of imbalance often present in real-life multilingual classification datasets: an uneven distribution of labels across languages. We show evidence that fine-tuning a transformer-based Large Language Model (LLM) on a dataset with this imbalance leads to worse performance, a more pronounced separation of languages in the latent space, and the promotion of uninformative features. We modify the traditional class weighing approach to imbalance by calculating class weights separately for each language and show that this helps mitigate those detrimental effects. These results create awareness of the negative effects of language-specific class imbalance in multilingual fine-tuning and the way in which the model learns to rely on the separation of languages to perform the task.