BanglaIPA: Towards Robust Text-to-IPA Transcription with Contextual Rewriting in Bengali

2026-01-05

arXiv.org

AI Summary

This study addresses the lack of a robust automatic IPA transcription system for Bengali capable of handling standard language, dialectal variation, numerals, and out-of-vocabulary words. The authors propose a novel context-aware rewriting mechanism that integrates character-level modeling with word-level alignment, marking the first approach to jointly leverage both character- and word-level information for high-accuracy, multi-dialect IPA transcription. By precomputing word-to-IPA mappings, the method significantly enhances inference efficiency while effectively supporting numeral transcription and generalization to unseen words. Evaluated on the DUAL-IPA dataset, the approach achieves an average word error rate of 11.4%, representing a substantial relative reduction of 58.4%–78.7% over existing baselines, thereby establishing a notable advance in both robustness and computational efficiency.

Technology Category

• Natural Language Processing: Conversational AI/Dialog Systems
• Machine Learning: Large Multimodal Models (LMMs)
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Despite its widespread use, Bengali lacks a robust automated International Phonetic Alphabet (IPA) transcription system that effectively supports both standard language and regional dialectal texts. Existing approaches struggle to handle regional variations, numerical expressions, and generalize poorly to previously unseen words. To address these limitations, we propose BanglaIPA, a novel IPA generation system that integrates a character-based vocabulary with word-level alignment. The proposed system accurately handles Bengali numerals and demonstrates strong performance across regional dialects. BanglaIPA improves inference efficiency by leveraging a precomputed word-to-IPA mapping dictionary for previously observed words. The system is evaluated on the standard Bengali and six regional variations of the DUAL-IPA dataset. Experimental results show that BanglaIPA outperforms baseline IPA transcription models by 58.4-78.7% and achieves an overall mean word error rate of 11.4%, highlighting its robustness in phonetic transcription generation for the Bengali language.

Problem

• Bengali
• IPA transcription
• regional dialects
• numerical expressions
• out-of-vocabulary words

Innovation

• IPA transcription
• Bengali dialects
• character-based vocabulary
• word-level alignment
• precomputed mapping