Tokenization and Morphology in Multilingual Language Models: A Comparative Analysis of mT5 and ByT5

TL;DR

This paper compares how different tokenization strategies in multilingual models affect their encoding of morphological knowledge across various languages, revealing insights into layer-specific encoding and data benefits.

Contribution

It provides a comparative analysis of subword and character-level tokenization in multilingual models, highlighting their impact on morphological understanding.

Findings

Models encode morphology differently across languages.

Morphological knowledge is concentrated in middle and late layers.

Languages with irregular morphology benefit from more pre-training data.

Abstract

Morphology is a crucial factor for multilingual language modeling as it poses direct challenges for tokenization. Here, we seek to understand how tokenization influences the morphological knowledge encoded in multilingual language models. Specifically, we capture the impact of tokenization by contrasting two multilingual language models: mT5 and ByT5. The two models share the same architecture, training objective, and training data and only differ in their tokenization strategies: subword tokenization vs.\@ character-level tokenization. Probing the morphological knowledge encoded in these models on four tasks and 17 languages, our analyses show that the models learn the morphological systems of some languages better than others and that morphological information is encoded in the middle and late layers. Finally, we show that languages with more irregularities benefit more from having a higher share of the pre-training data.