At this week’s lab meeting, Logan will discuss tokenization-free approaches to NLP. We will focus on CANINE (Clark et al. 2022):
Abstract: Pipelined NLP systems have largely been superseded by end-to-end neural modeling, yet nearly all commonly-used models still require an explicit tokenization step. While recent tokenization approaches based on data-derived subword lexicons are less brittle than manually engineered tokenizers, these techniques are not equally suited to all languages, and the use of any fixed vocabulary may limit a model’s ability to adapt. In this paper, we present CANINE, a neural encoder that operates directly on character sequences, without explicit tokenization or vocabulary, and a pre-training strategy that operates either directly on characters or optionally uses subwords as a soft inductive bias. To use its finer-grained input effectively and efficiently, CANINE combines downsampling, which reduces the input sequence length, with a deep transformer stack, which encodes context. CANINE outperforms a comparable mBERT model by 2.8 F1 on TyDi QA, a challenging multilingual benchmark, despite having 28% fewer model parameters.
Time permitting, we may also discuss the Perceiver IO architecture which offers alternative means for avoiding tokenization.
Wednesday, 24 May at 12pm
This will be a hybrid meeting at TASC1 9408. The zoom link will be posted on zulip.
