Biomedical Language Models are Robust to Sub-optimal Tokenization

TL;DR

This study investigates whether using more accurate biomedical tokenizers improves language model performance on biomedical NLP tasks, finding that pre-training is robust even with sub-optimal tokenization.

Contribution

It demonstrates that biomedical language models are surprisingly unaffected by sub-optimal tokenization during pre-training, challenging assumptions about the importance of tokenizer accuracy.

Findings

Standard biomedical tokenizers often fail to segment terms into meaningful parts.

Using a more accurate biomedical tokenizer does not significantly improve model performance.

Biomedical pre-training is robust to sub-optimal tokenization, as shown by various evaluation metrics.

Abstract

As opposed to general English, many concepts in biomedical terminology have been designed in recent history by biomedical professionals with the goal of being precise and concise. This is often achieved by concatenating meaningful biomedical morphemes to create new semantic units. Nevertheless, most modern biomedical language models (LMs) are pre-trained using standard domain-specific tokenizers derived from large scale biomedical corpus statistics without explicitly leveraging the agglutinating nature of biomedical language. In this work, we first find that standard open-domain and biomedical tokenizers are largely unable to segment biomedical terms into meaningful components. Therefore, we hypothesize that using a tokenizer which segments biomedical terminology more accurately would enable biomedical LMs to improve their performance on downstream biomedical NLP tasks, especially ones which involve biomedical terms directly such as named entity recognition (NER) and entity linking. Surprisingly, we find that pre-training a biomedical LM using a more accurate biomedical tokenizer does not improve the entity representation quality of a language model as measured by several intrinsic and extrinsic measures such as masked language modeling prediction (MLM) accuracy as well as NER and entity linking performance. These quantitative findings, along with a case study which explores entity representation quality more directly, suggest that the biomedical pre-training process is quite robust to instances of sub-optimal tokenization.