MUTEX: Leveraging Multilingual Transformers and Conditional Random Fields for Enhanced Urdu Toxic Span Detection

TL;DR

This paper introduces MUTEX, a multilingual transformer and CRF-based framework that significantly improves Urdu toxic span detection by leveraging token-level annotated data, addressing linguistic complexities, and outperforming previous methods.

Contribution

MUTEX is the first supervised model for Urdu toxic span detection using a multilingual transformer and CRF, enhancing interpretability and handling linguistic challenges.

Findings

Achieved 60% token-level F1 score on Urdu toxic span detection.

Transformer models better capture contextual toxicity and linguistic variations.

MUTEX outperforms existing approaches in fine-grained toxic span detection.

Abstract

Urdu toxic span detection remains limited because most existing systems rely on sentence-level classification and fail to identify the specific toxic spans within those text. It is further exacerbated by the multiple factors i.e. lack of token-level annotated resources, linguistic complexity of Urdu, frequent code-switching, informal expressions, and rich morphological variations. In this research, we propose MUTEX: a multilingual transformer combined with conditional random fields (CRF) for Urdu toxic span detection framework that uses manually annotated token-level toxic span dataset to improve performance and interpretability. MUTEX uses XLM RoBERTa with CRF layer to perform sequence labeling and is tested on multi-domain data extracted from social media, online news, and YouTube reviews using token-level F1 to evaluate fine-grained span detection. The results indicate that MUTEX achieves 60% token-level F1 score that is the first supervised baseline for Urdu toxic span detection. Further examination reveals that transformer-based models are more effective at implicitly capturing the contextual toxicity and are able to address the issues of code-switching and morphological variation than other models.