A Shared Attention Mechanism for Interpretation of Neural Automatic Post-Editing Systems

机译：神经自动后编辑系统解释的共享注意机制

获取原文

获取原文并翻译 | 示例

页面导航

摘要
著录项
相似文献
相关主题

摘要

Automatic post-editing (APE) systems aim to correct the systematic errors made by machine translators. In this paper, we propose a neural APE system that encodes the source (src) and machine translated (mt) sentences with two separate encoders, but leverages a shared attention mechanism to better understand how the two inputs contribute to the generation of the post-edited (pe) sentences. Our empirical observations have showed that when the mt is incorrect, the attention shifts weight toward tokens in the src sentence to properly edit the incorrect translation. The model has been trained and evaluated on the official data from the WMT16 and WMT17 APE IT domain English-German shared tasks. Additionally, we have used the extra 500K artificial data provided by the shared task. Our system has been able to reproduce the accuracies of systems trained with the same data, while at the same time providing better interpretability.

机译：自动后期编辑（APE）系统旨在纠正机器翻译人员所犯的系统错误。在本文中，我们提出了一种神经APE系统，该系统使用两个单独的编码器对源（src）和机器翻译（mt）句子进行编码，但利用共享的关注机制更好地了解这两个输入如何有助于生成后置编辑（体育）句子。我们的经验观察表明，当mt错误时，注意力会转移到src句子中的标记上，以正确编辑错误的翻译。该模型已根据来自WMT16和WMT17 APE IT领域英语-德语共享任务的官方数据进行了培训和评估。此外，我们还使用了共享任务提供的额外500K人工数据。我们的系统能够重现使用相同数据训练的系统的准确性，同时提供更好的解释性。

著录项

来源
《Second workshop on neural machine translation 2018》|2018年|11-17|共7页
会议地点 Melbourne(AU)
作者
Inigo Jauregi Unanue; Ehsan Zare Borzeshi; Massimo Piccardi;
展开▼
作者单位

University of Technology Sydney, Sydney, Australia,Capital Markets Cooperative Research Centre, Sydney, Australia;

Capital Markets Cooperative Research Centre, Sydney, Australia;

University of Technology Sydney, Sydney, Australia;

展开▼
会议组织
原文格式 PDF
正文语种 eng
中图分类
关键词

相似文献

外文文献
中文文献
专利

1. Multistep Flow Prediction on Car-Sharing Systems: A Multi-Graph Convolutional Neural Network with Attention Mechanism [J] . Hongming Zhu, Yi Luo, Qin Liu, International journal of software engineering and knowledge engineering . 2019,第11a12期

机译：汽车共享系统的多步流量预测：具有注意机制的多图卷积神经网络
2. Commonalities in the neural mechanisms underlying automatic attentional shifts by gaze, gestures, and symbols. [J] . Sato W, Kochiyama T, Uono S, NeuroImage . 2009,第3期

机译：通过注视，手势和符号进行自动注意力转移的神经机制的共性。
3. Automatic Versus Contingent Mechanisms of Sensory-Driven Neural Biasing and Reflexive Attention [J] . Joseph B. Hopfinger, Anthony J. Ries Journal of Cognitive Neuroscience . 2005,第8期

机译：感官驱动的神经性偏向和反身注意的自动对然机制
4. A Shared Attention Mechanism for Interpretation of Neural Automatic Post-Editing Systems [C] . Inigo Jauregi Unanue, Ehsan Zare Borzeshi, Massimo Piccardi Annual meeting of the Association for Computational Linguistics . 2018

机译：诠释神经自动后编辑系统的共同关注机制
5. Neural Mechanisms of Spatial Attention: Electrophysiological Studies of Attentional Feedback Priming in Human Primary Visual Cortex [D] . Royston, Ashley. 2018

机译：空间关注的神经机制：人原发性视觉皮层中注意力反馈引发的电生理学研究
6. Automatic diagnosis of macular diseases from OCT volume based on its two-dimensional feature map and convolutional neural network with attention mechanism [O] . Yankui Sun, Haoran Zhang, Xianlin Yao 2020

机译：基于其二维特征图和卷积神经网络自动诊断来自OCT卷的关注机制
7. A Shared Attention Mechanism for Interpretation of Neural Automatic Post-Editing Systems [O] . Inigo Jauregi Unanue, Ehsan Zare Borzeshi, Massimo Piccardi 2018

机译：诠释神经自动后编辑系统的共同关注机制

A Shared Attention Mechanism for Interpretation of Neural Automatic Post-Editing Systems

摘要

著录项

相似文献

相关主题

期刊订阅