One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography

• URL: http://arxiv.org/abs/2608.25904v2
• Date: Thu, 27 Aug 2026 20:32:02 GMT
• Title: One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography
• Abstract summary: We systematically compare different input representations in autoregressive multilingual pretraining, comparing orthographic text, IPA, and romanization.<n>Romanization yields the strongest cross-lingual transfer, and the advantage over text widens with scale.<n>Our results indicate that for multilingual models spanning typologically diverse scripts, romanization should be treated as a core design choice applied at pretraining.
• Score: 35.26843528120082
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multilingual language models transfer knowledge across languages through shared subword vocabulary, a mechanism that breaks down when related languages use different writing systems. Prior work addresses this via script equalization (romanization or IPA transcription), but direct comparisons are rare; the focus has been on encoder-only models, with most work adapting existing pretrained models. We systematically compare different input representations in autoregressive multilingual pretraining, comparing orthographic text, IPA, and romanization in a controlled setup across three scales (467M, 709M, and 1.03B) on eight languages in four typologically motivated pairs. Across a wide range of downstream tasks on seen and unseen languages, romanized pretraining yields the strongest cross-lingual transfer, and the advantage over text widens with scale. IPA improves over text in most settings but trails romanization. Surprisingly, finetuning a text-pretrained model on romanized data hurts performance on languages already covered by the base model, only marginally helping when the model lacks script coverage. Our results indicate that for multilingual models spanning typologically diverse scripts, to obtain maximum benefits, romanization should be treated as a core design choice applied at pretraining rather than a post hoc fix.

Related papers

• The Curse of Multilinguality in Lexical Normalization [0.0]
  A popular shortcut is to train a single model on many languages at once.<n>We ask how many languages should such a model be trained on?<n>We find a clear curse of multilinguality: accuracy is highest when a language is trained with only a few others.
  arXiv  Detail & Related papers  (2026-08-27T22:48:02Z)
• A two-stage transliteration approach to improve performance of a multilingual ASR [1.9511556030544333]
  This paper presents an approach to build a language-agnostic end-to-end model trained on a grapheme set. We performed experiments with an end-to-end multilingual speech recognition system for two Indic languages.
  arXiv  Detail & Related papers  (2024-10-09T05:30:33Z)
• Breaking the Script Barrier in Multilingual Pre-Trained Language Models with Transliteration-Based Post-Training Alignment [50.27950279695363]
  The transfer performance is often hindered when a low-resource target language is written in a different script than the high-resource source language. Inspired by recent work that uses transliteration to address this problem, our paper proposes a transliteration-based post-pretraining alignment (PPA) method.
  arXiv  Detail & Related papers  (2024-06-28T08:59:24Z)
• TransliCo: A Contrastive Learning Framework to Address the Script Barrier in Multilingual Pretrained Language Models [50.40191599304911]
  We propose TransliCo to fine-tune an mPLM by contrasting sentences in its training data and their transliterations in a unified script. We show that Furina outperforms the original Glot500-m on various zero-shot crosslingual transfer tasks.
  arXiv  Detail & Related papers  (2024-01-12T15:12:48Z)
• Soft Language Clustering for Multilingual Model Pre-training [57.18058739931463]
  We propose XLM-P, which contextually retrieves prompts as flexible guidance for encoding instances conditionally. Our XLM-P enables (1) lightweight modeling of language-invariant and language-specific knowledge across languages, and (2) easy integration with other multilingual pre-training methods.
  arXiv  Detail & Related papers  (2023-06-13T08:08:08Z)
• UNKs Everywhere: Adapting Multilingual Language Models to New Scripts [103.79021395138423]
  Massively multilingual language models such as multilingual BERT (mBERT) and XLM-R offer state-of-the-art cross-lingual transfer performance on a range of NLP tasks. Due to their limited capacity and large differences in pretraining data, there is a profound performance gap between resource-rich and resource-poor target languages. We propose novel data-efficient methods that enable quick and effective adaptation of pretrained multilingual models to such low-resource languages and unseen scripts.
  arXiv  Detail & Related papers  (2020-12-31T11:37:28Z)
• Multilingual Translation with Extensible Multilingual Pretraining and Finetuning [77.33262578776291]
  Previous work has demonstrated that machine translation systems can be created by finetuning on bitext. We show that multilingual translation models can be created through multilingual finetuning. We demonstrate that pretrained models can be extended to incorporate additional languages without loss of performance.
  arXiv  Detail & Related papers  (2020-08-02T05:36:55Z)
• InfoXLM: An Information-Theoretic Framework for Cross-Lingual Language Model Pre-Training [135.12061144759517]
  We present an information-theoretic framework that formulates cross-lingual language model pre-training. We propose a new pre-training task based on contrastive learning. By leveraging both monolingual and parallel corpora, we jointly train the pretext to improve the cross-lingual transferability of pre-trained models.
  arXiv  Detail & Related papers  (2020-07-15T16:58:01Z)
• Cross-lingual Transfer of Sentiment Classifiers [2.1600185911839893]
  Cross-lingual word embeddings transform vector spaces of different languages so that similar words are aligned. We use cross-lingual embeddings to transfer machine learning prediction models for Twitter sentiment between 13 languages.
  arXiv  Detail & Related papers  (2020-05-15T10:15:27Z)
• Language-agnostic Multilingual Modeling [23.06484126933893]
  We build a language-agnostic multilingual ASR system which transforms all languages to one writing system through a many-to-one transliteration transducer. We show with four Indic languages, namely, Hindi, Bengali, Tamil and Kannada, that the language-agnostic multilingual model achieves up to 10% relative reduction in Word Error Rate (WER) over a language-dependent multilingual model.
  arXiv  Detail & Related papers  (2020-04-20T18:57:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.