DiSAN: Directional Self-Attention Network for RNN/CNN-free Language Understanding

Recurrent neural nets (RNN) and convolutional neural nets (CNN) are widely used in NLP tasks to capture the long-term and local dependencies respectively. Attention mechanisms have recently attracted enormous interest due to their highly parallelizable computation, significantly less training time, and flexibility in modeling dependencies. We propose a novel attention mechanism in which the attention between elements from input sequence(s) is directional and multi-dimensional, i.e., feature-wise. A light-weight neural net, "Directional Self-Attention Network (DiSAN)", is then proposed to learn sentence embedding, based solely on the proposed attention without any RNN/CNN structure. DiSAN is only composed of a directional self-attention block with temporal order encoded, followed by a multi-dimensional attention that compresses the sequence into a vector representation. Despite this simple form, DiSAN outperforms complicated RNN/CNN models on both prediction quality and efficiency. It achieves the best test accuracy among all sentence encoding methods and improves the most recent best result by about 1.0% on the Stanford Natural Language Inference (SNLI) dataset, and shows the state-of-the-art test accuracy on the Stanford Sentiment Treebank (SST), Sentences Involving Compositional Knowledge (SICK), TREC Question-type Classification and Multi-Genre Natural Language Inference (MultiNLI) datasets.