LionGuard: Building a Contextualized Moderation Classifier to Tackle Localized Unsafe Content

TL;DR

LionGuard is a Singapore-specific moderation classifier that improves safety detection for local languages like Singlish, outperforming generic APIs and emphasizing the importance of localization in moderation tools.

Contribution

We introduce LionGuard, a localized moderation classifier tailored for Singaporean context, demonstrating significant performance gains over existing non-localized moderation APIs.

Findings

LionGuard outperforms existing APIs by 14-51% on Singlish data.

Localization enhances moderation accuracy for low-resource languages.

The approach is practical and scalable for diverse language contexts.

Abstract

As large language models (LLMs) become increasingly prevalent in a wide variety of applications, concerns about the safety of their outputs have become more significant. Most efforts at safety-tuning or moderation today take on a predominantly Western-centric view of safety, especially for toxic, hateful, or violent speech. In this paper, we describe LionGuard, a Singapore-contextualized moderation classifier that can serve as guardrails against unsafe LLM outputs. When assessed on Singlish data, LionGuard outperforms existing widely-used moderation APIs, which are not finetuned for the Singapore context, by 14% (binary) and up to 51% (multi-label). Our work highlights the benefits of localization for moderation classifiers and presents a practical and scalable approach for low-resource languages.