Generating Diverse Translation by Manipulating Multi-Head Attention
arXiv:1911.09333
Abstract
Transformer model has been widely used on machine translation tasks and obtained state-of-the-art results. In this paper, we report an interesting phenomenon in its encoder-decoder multi-head attention: different attention heads of the final decoder layer align to different word translation candidates. We empirically verify this discovery and propose a method to generate diverse translations by manipulating heads. Furthermore, we make use of these diverse translations with the back-translation technique for better data augmentation. Experiment results show that our method generates diverse translations without severe drop in translation quality. Experiments also show that back-translation with these diverse translations could bring significant improvement on performance on translation tasks. An auxiliary experiment of conversation response generation task proves the effect of diversity as well.
Accepted by AAAI 2020
References in corpus (14)
- Convolutional Sequence to Sequence Learning
- Effective Approaches to Attention-based Neural Machine Translation
- A Simple, Fast Diverse Decoding Algorithm for Neural Generation
- Neural Responding Machine for Short-Text Conversation
- A Deep Generative Framework for Paraphrase Generation
- Investigating Backtranslation in Neural Machine Translation
- Understanding Back-Translation at Scale
- Analyzing Uncertainty in Neural Machine Translation
- Creativity: Generating Diverse Questions using Variational Autoencoders
- D-PAGE: Diverse Paraphrase Generation
- Multi-Head Attention with Disagreement Regularization
- Convolutional Self-Attention Networks
- Information Aggregation for Multi-Head Attention with Routing-by-Agreement
- Sequence to Sequence Mixture Model for Diverse Machine Translation