时 间：2022/10/19 下午19:00
Language models are defined over a finite set of inputs, which creates a vocabulary bottleneck when we attempt to scale the number of supported languages. Tackling this bottleneck results in a trade-off between what can be represented in the embedding matrix and computational issues in the output layer. This paper introduces PIXEL, the Pixel-based Encoder of Language, which suffers from neither of these issues. PIXEL is a pretrained language model that renders text as images, making it possible to transfer representations across languages based on orthographic similarity or the co-activation of pixels. PIXEL is trained to reconstruct the pixels of masked patches, instead of predicting a distribution over tokens. We pretrain the 86M parameter PIXEL model on the same English data as BERT and evaluate on syntactic and semantic tasks in typologically diverse languages, including various non-Latin scripts. We find that PIXEL substantially outperforms BERT on syntactic and semantic processing tasks on scripts that are not found in the pretraining data, but PIXEL is slightly weaker than BERT when working with Latin scripts. Furthermore, we find that PIXEL is more robust to noisy text inputs than BERT, further confirming the benefits of modelling language with pixels.
Desmond Elliott is an Assistant Professor at the University of Copenhagen where he builds and attempts to understand multimodal and multilingual models. His work received the Best Long Paper Award at EMNLP 2021 and an Area Chair Favourite paper at COLING 2018. He co-organised the Multimodal Machine Translation Shared Task from 2016–2018, the 2018 Frederick Jelinek Memorial Workshop on Grounded Sequence-to-Sequence Learning, the How2 Challenge Workshop at ICML 2019, and the Workshop on Multilingual Multimodal Learning at ACL 2022.