SiNFluD: Creating and Evaluating Figurative Language Dataset for Sindhi

TL;DR

This paper introduces SiNFluD, a new Sindhi figurative language dataset, and evaluates multiple models for classification, with XLM-RoBERTa-XL performing best.

Contribution

The creation of the first benchmark dataset for Sindhi figurative language classification and baseline evaluations of several models.

Findings

XLM-RoBERTa-XL achieves the highest accuracy among tested models.

Inter-annotator agreement of 0.81 indicates reliable labeling.

Baseline results establish a foundation for future Sindhi NLP research.

Abstract

In this article, we introduce SiNFluD, a novel benchmark dataset for Sindhi figurative language classification. We first collect raw text from various blogs, social media platforms, and literary sources, and subsequently prepare the corpus for annotation. Two native annotators label the data using the Doccano text annotation tool, achieving an inter-annotator agreement of 0.81. We then establish baseline results using 5-fold and 10-fold cross-validation. Finally, we evaluate mBERT, XLM-RoBERTa, and XLM-RoBERTa-XL models, along with SetFit for few-shot fine-tuning of sentence transformers. Among these, the pretrained XLM-RoBERTa-XL achieves the best performance.