# Char2Subword: Extending the Subword Embedding Space Using Robust Character Compositionality

arXiv: https://arxiv.org/abs/2010.12730 · HTML: https://mlsanity.com/paper/2010.12730/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Oct 2020.

## Built on by (most cited first, 11 of 11)

- [Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLP](https://mlsanity.com/paper/2112.10508/) (2021)
- [Language Modelling with Pixels](https://mlsanity.com/paper/2207.06991/) (2022)
- [What do tokens know about their characters and how do they know it?](https://mlsanity.com/paper/2206.02608/) (2022)
- [The Decades Progress on Code-Switching Research in NLP: A Systematic Survey on Trends and Challenges](https://mlsanity.com/paper/2212.09660/) (2022)
- [RoBERTuito: a pre-trained language model for social media text in Spanish](https://mlsanity.com/paper/2111.09453/) (2021)
- [Inducing Character-level Structure in Subword-based Language Models with Type-level Interchange Intervention Training](https://mlsanity.com/paper/2212.09897/) (2022)
- [A Vocabulary-Free Multilingual Neural Tokenizer for End-to-End Task Learning](https://mlsanity.com/paper/2204.10815/) (2022)
- [Multilingual Transfer Learning for Code-Switched Language and Speech Neural Modeling](https://mlsanity.com/paper/2104.06268/) (2021)
- [From Characters to Words: Hierarchical Pre-trained Language Model for Open-vocabulary Language Understanding](https://mlsanity.com/paper/2305.14571/) (2023)
- [Word-Level Representation From Bytes For Language Modeling](https://mlsanity.com/paper/2211.12677/) (2022)
- [SCRIPT: A Subcharacter Compositional Representation Injection Module for Korean Pre-Trained Language Models](https://mlsanity.com/paper/2604.12377/) (2026)

## Abstract (arXiv)

Byte-pair encoding (BPE) is a ubiquitous algorithm in the subword tokenization process of language models as it provides multiple benefits. However, this process is solely based on pre-training data statistics, making it hard for the tokenizer to handle infrequent spellings. On the other hand, though robust to misspellings, pure character-level models often lead to unreasonably long sequences and make it harder for the model to learn meaningful words. To alleviate these challenges, we propose a character-based subword module (char2subword) that learns the subword embedding table in pre-trained models like BERT. Our char2subword module builds representations from characters out of the subword vocabulary, and it can be used as a drop-in replacement of the subword embedding table. The module is robust to character-level alterations such as misspellings, word inflection, casing, and punctuation. We integrate it further with BERT through pre-training while keeping BERT transformer parameters fixed--and thus, providing a practical method. Finally, we show that incorporating our module to mBERT significantly improves the performance on the social media linguistic code-switching evaluation (LinCE) benchmark.