Neural Generative Question Answering
Neural Machine Translation of Rare Words with Subword Units
将BPE用于处理翻译系统中的OOV问题．
Convolutional Sequence to Sequence Learning
facebook发布的给予CNN的NMT系统．
Attention Is All You Need
google发布的完全给予attention的NMT系统．可以并行．
Minimum Risk Training for Neural Machine Translation
改变了训练准则为最小风险训练
Sequence-to-Sequence Learning as Beam-Search Optimization
加入了beam-search
Sequence Level Training with Recurrent Neural Networks
因为在训练和预测时候的解码过程不同，导致了decoder的输出分布不同，这个文章就是研究这个问题．
Dual Learning for Machine Translation
基于对偶学习的NMT．
A Diversity-Promoting Objective Function for Neural Conversation Models
在对话系统中容易产生通用回复，文章加入了用问题和答案的忽信息来优化模型，使得通用回答降少．
Deep Reinforcement Learning for Dialogue Generation
论文提出了使用强化学习来优化对话系统
Adversarial Learning for Neural Dialogue Generation
论文提出了使用Gan来实现对话系统
DocChat: An Information Retrieval Approach for Chatbot Engines Using Unstructured Documents
来自微软亚洲研究院自动聊天机器人方面的研究