Guardians of Discourse: Evaluating LLMs on Multilingual Offensive Language Detection

TL;DR

This paper evaluates the effectiveness of large language models in detecting offensive language across English, Spanish, and German, highlighting the influence of prompt language, translation data, and inherent biases.

Contribution

First comprehensive multilingual evaluation of LLMs for offensive language detection, analyzing prompt language effects and dataset biases.

Findings

LLMs perform variably across languages and settings.

Prompt language influences detection accuracy.

Biases in models and datasets affect mispredictions.

Abstract

Identifying offensive language is essential for maintaining safety and sustainability in the social media era. Though large language models (LLMs) have demonstrated encouraging potential in social media analytics, they lack thorough evaluation when in offensive language detection, particularly in multilingual environments. We for the first time evaluate multilingual offensive language detection of LLMs in three languages: English, Spanish, and German with three LLMs, GPT-3.5, Flan-T5, and Mistral, in both monolingual and multilingual settings. We further examine the impact of different prompt languages and augmented translation data for the task in non-English contexts. Furthermore, we discuss the impact of the inherent bias in LLMs and the datasets in the mispredictions related to sensitive topics.