When Scripts Diverge: Strengthening Low-Resource Neural Machine Translation Through Phonetic Cross-Lingual Transfer

Multilingual Neural Machine Translation (MNMT) models enhance translation quality for low-resource languages by exploiting cross-lingual similarities during training—a process known as knowledge transfer. This transfer is particularly effective between languages that share lexical or structural features, often enabled by a common orthography. However, languages with strong phonetic and lexical similarities but distinct writing systems experience limited benefits, as the absence of a shared orthography hinders knowledge transfer. To address this limitation, we propose an approach based on phonetic information that enhances token-level alignment across scripts by leveraging transliterations. We systematically evaluate several phonetic transcription techniques and strategies for incorporating phonetic information into NMT models. Our results show that using a shared encoder to process orthographic and phonetic inputs separately consistently yields the best performance for Khmer, Thai, and Lao in both directions with English, and that our custom Cognate-Aware Transliteration (CAT) method consistently improves translation quality over the baseline.

• Publication year

  2025

• Venue

  Proceedings of the 5th Workshop on Multilingual Representation Learning (MRL 2025)

• Publication date

  Unknown publication date

• Fields of study

  Not labeled

• Identifiers
  DOI 10.18653/v1/2025.mrl-main.22
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Prompting with Phonemes: Enhancing LLM Multilinguality for non-Latin Script Languages

  2024cited by this paper
• Scaling neural machine translation to 200 languages

  2024cited by this paper
• CORI: CJKV Benchmark with Romanization Integration - a Step towards Cross-lingual Transfer beyond Textual Scripts

  2024cited by this paper
• RomanSetu: Efficiently unlocking multilingual capabilities of Large Language Models via Romanization

  2024cited by this paper
• Neural Methods for Aligning Large-Scale Parallel Corpora from the Web for South and East Asian Languages

  2024cited by this paper
• IndicTrans2: Towards High-Quality and Accessible Machine Translation Models for all 22 Scheduled Indian Languages

  2023cited by this paper
• Romanization-based Large-scale Adaptation of Multilingual Language Models

  2023cited by this paper
• Multilingual Pixel Representations for Translation and Effective Cross-lingual Transfer

  2023cited by this paper
• PyThaiNLP: Thai Natural Language Processing in Python

  2023cited by this paper
• Enhancing Cross-lingual Transfer via Phonemic Transcription Integration

  2023cited by this paper
• African Substrates Rather Than European Lexifiers to Augment African-diaspora Creole Translation

  2023cited by this paper
• Tokenization Impacts Multilingual Language Modeling: Assessing Vocabulary Allocation and Overlap Across Languages

  2023cited by this paper
• Does Transliteration Help Multilingual Language Modeling?

  2022cited by this paper
• Nick J. Enfield, Mainland Southeast Asian Languages, A Concise Typological Introduction

  2022cited by this paper
• Overlap-based Vocabulary Generation Improves Cross-lingual Transfer Among Related Languages

  2022cited by this paper
• Machine Translation of Low-Resource Indo-European Languages

  2021cited by this paper
• When Being Unseen from mBERT is just the Beginning: Handling New Languages With Multilingual Language Models

  2020cited by this paper
• Complete Multilingual Neural Machine Translation

  2020cited by this paper
• Improving Multilingual Neural Machine Translation For Low-Resource Languages: French, English - Vietnamese

  2020cited by this paper
• BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension

  2019cited by this paper
• A Universal Parent Model for Low-Resource Neural Machine Translation Transfer

  2019cited by this paper
• Beto, Bentz, Becas: The Surprising Cross-Lingual Effectiveness of BERT

  2019cited by this paper
• Massively Multilingual Neural Machine Translation

  2019cited by this paper
• Unsupervised Cross-lingual Representation Learning at Scale

  2019cited by this paper
• How Multilingual is Multilingual BERT?

  2019cited by this paper
• Cognate-aware morphological segmentation for multilingual neural translation

  2018cited by this paper
• Multi-Source Cross-Lingual Model Transfer: Learning What to Share

  2018cited by this paper
• A Call for Clarity in Reporting BLEU Scores

  2018cited by this paper
• Out-of-the-box Universal Romanization Tool uroman

  2018cited by this paper
• Rapid Adaptation of Neural Machine Translation to New Languages

  2018cited by this paper
• chrF++: words helping character n-grams

  2017cited by this paper
• Attention is All you Need

  2017cited by this paper
• Transfer Learning for Low-Resource Neural Machine Translation

  2016cited by this paper
• Sound Correspondences in the World's Languages

  2013cited by this paper
• Bleu: a Method for Automatic Evaluation of Machine Translation

  2002cited by this paper
• Etude comparative de la distribution florale dans une portion des Alpes et des Jura

  year unknowncited by this paper

• No citing papers are available for this paper.