SinFoS: A Parallel Dataset for Translating Sinhala Figures of Speech

2026-02-09

AI Summary

This study addresses the challenges of translating idiomatic expressions in low-resource languages like Sinhala, where data scarcity and cultural specificity hinder neural machine translation and prevent large language models from accurately conveying cultural nuances. The authors present the first systematically constructed and publicly released parallel corpus comprising 2,344 Sinhala rhetorical expressions, annotated with their cultural origins and cross-lingual equivalents. Leveraging this resource, they train a binary classifier to distinguish rhetorical types, achieving 92% accuracy, and evaluate the performance of mainstream large language models on this task. Their findings reveal significant limitations in current models when handling culturally sensitive expressions in low-resource settings, thereby establishing a new benchmark and providing essential data support for advancing machine translation of such linguistically and culturally complex phenomena.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Knowledge Representation Languages

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Figures of Speech (FoS) consist of multi-word phrases that are deeply intertwined with culture. While Neural Machine Translation (NMT) performs relatively well with the figurative expressions of high-resource languages, it often faces challenges when dealing with low-resource languages like Sinhala due to limited available data. To address this limitation, we introduce a corpus of 2,344 Sinhala figures of speech with cultural and cross-lingual annotations. We examine this dataset to classify the cultural origins of the figures of speech and to identify their cross-lingual equivalents. Additionally, we have developed a binary classifier to differentiate between two types of FOS in the dataset, achieving an accuracy rate of approximately 92%. We also evaluate the performance of existing LLMs on this dataset. Our findings reveal significant shortcomings in the current capabilities of LLMs, as these models often struggle to accurately convey idiomatic meanings. By making this dataset publicly available, we offer a crucial benchmark for future research in low-resource NLP and culturally aware machine translation.

Problem

• Figures of Speech
• Low-resource languages
• Neural Machine Translation
• Culturally aware translation
• Sinhala

Innovation

• Figures of Speech
• Low-resource NLP
• Culturally-aware Machine Translation
• Parallel Corpus
• Cross-lingual Annotation