Charformer: Fast Character Transformers via Gradient-based Subword Tokenization

TL;DR

Charformer introduces an end-to-end trainable model with a gradient-based subword tokenization module, enabling faster and more adaptable NLP models that operate directly on characters or bytes, outperforming traditional subword methods.

Contribution

The paper presents GBST, a novel differentiable subword tokenization method integrated into a Transformer, allowing end-to-end training and improved speed and flexibility over existing models.

Findings

Charformer outperforms byte-level baselines on multiple NLP tasks.

It improves Transformer speed by 28%-100%.

Maintains competitive accuracy with subword-based models.

Abstract

State-of-the-art models in natural language processing rely on separate rigid subword tokenization algorithms, which limit their generalization ability and adaptation to new settings. In this paper, we propose a new model inductive bias that learns a subword tokenization end-to-end as part of the model. To this end, we introduce a soft gradient-based subword tokenization module (GBST) that automatically learns latent subword representations from characters in a data-driven fashion. Concretely, GBST enumerates candidate subword blocks and learns to score them in a position-wise fashion using a block scoring network. We additionally introduce Charformer, a deep Transformer model that integrates GBST and operates on the byte level. Via extensive experiments on English GLUE, multilingual, and noisy text datasets, we show that Charformer outperforms a series of competitive byte-level baselines while generally performing on par and sometimes outperforming subword-based models. Additionally, Charformer is fast, improving the speed of both vanilla byte-level and subword-level Transformers by 28%-100% while maintaining competitive quality. We believe this work paves the way for highly performant token-free models that are trained completely end-to-end.