# The Curse of Multilinguality in Lexical Normalization

arXiv: https://arxiv.org/abs/2609.00329 · HTML: https://mlsanity.com/paper/2609.00329/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 14 references that are in the map.
- First posted on arXiv 27 Aug 2026.

## Builds on (most cited first, 8 of 8)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [Cross-lingual Language Model Pretraining](https://mlsanity.com/paper/1901.07291/) (2019)
- [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://mlsanity.com/paper/2105.13626/) (2021)
- [MAD-X: An Adapter-Based Framework for Multi-Task Cross-Lingual Transfer](https://mlsanity.com/paper/2005.00052/) (2020)
- [Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges](https://mlsanity.com/paper/1907.05019/) (2019)
- [MoNoise: Modeling Noise Using a Modular Normalization System](https://mlsanity.com/paper/1710.03476/) (2017)
- ['UFAL at MultiLexNorm 2021: Improving Multilingual Lexical Normalization by Fine-tuning ByT5](https://mlsanity.com/paper/2110.15248/) (2021)

## Abstract (arXiv)

Lexical normalization rewrites the noisy, non-standard words that fill user-generated text (tmrw, u, gr8) into their standard forms. Because labelled data is scarce for most languages, a popular shortcut is to train a single model on many languages at once. We ask a simple question: how many languages should such a model be trained on? Using one fixed-capacity character-level model and twelve languages from a standard benchmark, we vary the number of jointly trained languages from one to twelve and measure per-language accuracy. We find a clear curse of multilinguality: accuracy is highest when a language is trained with only a few others, often just one to four, and then falls steadily and substantially, dropping by about forty percent as the rest are piled on. A control that holds the total amount of training data constant makes the decline arrive sooner and fall further, which points to competition among the languages for one fixed-size model rather than to how much data is available. We also test whether a language's typological distance from the others predicts its ideal number of co-training languages, and find no dependable rule: any apparent relationship rests on a couple of languages and does not hold up. For compact normalization models, less can be more: a few languages beat pooling everything into a single model.