Models In a Spelling Bee: Language Models Implicitly Learn the Character Composition of Tokens
This addresses the problem of understanding how language models represent text for researchers, but it is incremental as it confirms implicit learning without practical gains.
The study found that pretrained language models like RoBERTa implicitly learn the character composition of tokens, accurately spelling up to a third of the vocabulary, but adding explicit character information did not improve language modeling performance.
Standard pretrained language models operate on sequences of subword tokens without direct access to the characters that compose each token's string representation. We probe the embedding layer of pretrained language models and show that models learn the internal character composition of whole word and subword tokens to a surprising extent, without ever seeing the characters coupled with the tokens. Our results show that the embedding layer of RoBERTa holds enough information to accurately spell up to a third of the vocabulary and reach high average character ngram overlap on all token types. We further test whether enriching subword models with additional character information can improve language modeling, and observe that this method has a near-identical learning curve as training without spelling-based enrichment. Overall, our results suggest that language modeling objectives incentivize the model to implicitly learn some notion of spelling, and that explicitly teaching the model how to spell does not appear to enhance its performance on such tasks.