QANet: Combining Local Convolution with Global Self-Attention for Reading Comprehension

TL;DR

QANet introduces a convolution and self-attention based architecture for reading comprehension that is significantly faster than RNN-based models while maintaining comparable accuracy, enabling larger data training.

Contribution

The paper presents QANet, a novel non-recurrent model combining convolution and self-attention, achieving faster training and inference without sacrificing accuracy.

Findings

QANet is 3-13x faster in training and 4-9x faster in inference than RNN models.

QANet achieves an F1 score of 84.6 on SQuAD with augmented data.

The model outperforms previous best results on SQuAD.

Abstract

Current end-to-end machine reading and question answering (Q\&A) models are primarily based on recurrent neural networks (RNNs) with attention. Despite their success, these models are often slow for both training and inference due to the sequential nature of RNNs. We propose a new Q\&A architecture called QANet, which does not require recurrent networks: Its encoder consists exclusively of convolution and self-attention, where convolution models local interactions and self-attention models global interactions. On the SQuAD dataset, our model is 3x to 13x faster in training and 4x to 9x faster in inference, while achieving equivalent accuracy to recurrent models. The speed-up gain allows us to train the model with much more data. We hence combine our model with data generated by backtranslation from a neural machine translation model. On the SQuAD dataset, our single model, trained with augmented data, achieves 84.6 F1 score on the test set, which is significantly better than the best published F1 score of 81.8.