Smart Reply
Published 8 August 2016Open access
Anjuli Kannan, Karol Kurach, Sujith Ravi, Tobias Kaufmann, Andrew Tomkins, Miklós Bálint
Citations183
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
This paper proposes and investigates a novel end-to-end method for automatically generating short email responses, called Smart Reply, which generates semantically diverse suggestions that can be used as complete email responses with just one tap on mobile.
Abstract
In this paper we propose and investigate a novel end-to-end method for automatically generating short email responses, called Smart Reply. It generates semantically diverse suggestions that can be used as complete email responses with just one tap on mobile. The system is currently used in Inbox by Gmail and is responsible for assisting with 10% of all mobile responses. It is designed to work at very high throughput and process hundreds of millions of messages daily. The system exploits state-of-the-art, large-scale deep learning.
Keywords
Computer ScienceDecision Sciences
Neural ComputationLong Short-Term Memory
98,079 Citations1997Sepp Hochreiter, Jürgen Schmidhuber
A novel, efficient, gradient based method called long short-term memory (LSTM) is introduced, which can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units.
Dropout: a simple way to prevent neural networks from overfitting
34,279 Citations2014Nitish Srivastava, Geoffrey E. Hinton +3 more
It is shown that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.
International Conference on Machine LearningRectified Linear Units Improve Restricted Boltzmann Machines
13,244 Citations2010Vinod Nair, Geoffrey E. Hinton
Restricted Boltzmann machines were developed using binary stochastic hidden units that learn features that are better for object recognition on the NORB dataset and face verification on the Labeled Faces in the Wild dataset.
arXiv (Cornell University)TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
9,778 Citations2016Martı́n Abadi, Ashish Agarwal +38 more
The TensorFlow interface and an implementation of that interface that is built at Google are described, which has been used for conducting research and for deploying machine learning systems into production across more than a dozen areas of computer science and other fields.
Adaptive Subgradient Methods for Online Learning and Stochastic Optimization
8,612 Citations2011John C. Duchi, Elad Hazan +1 more
Show and tell: A neural image caption generator
6,363 Citations2015Oriol Vinyals, Alexander Toshev +2 more
This paper presents a generative model based on a deep recurrent architecture that combines recent advances in computer vision and machine translation and that can be used to generate natural sentences describing an image.
arXiv (Cornell University)Sequence to Sequence Learning with Neural Networks
3,526 Citations2014Ilya Sutskever, Oriol Vinyals +1 more
This paper presents a general end-to-end approach to sequence learning that makes minimal assumptions on the sequence structure, and finds that reversing the order of the words in all source sentences improved the LSTM's performance markedly, because doing so introduced many short term dependencies between the source and the target sentence which made the optimization problem easier.
Semi-supervised learning using Gaussian fields and harmonic functions
3,377 Citations2003Xiaojin Zhu, Zoubin Ghahramani +1 more
Long short-term memory recurrent neural network architectures for large scale acoustic modeling
2,924 Citations2014Haşim Sak, Andrew Senior +1 more
The first distributed training of LSTM RNNs using asynchronous stochastic gradient descent optimization on a large cluster of machines is introduced and it is shown that a two-layer deep LSTm RNN where each L STM layer has a linear recurrent projection layer can exceed state-of-the-art speech recognition performance.
IEEE Transactions on Acoustics Speech and Signal ProcessingEstimation of probabilities from sparse data for the language model component of a speech recognizer
1,645 Citations1987Slava M. Katz
The model offers, via a nonlinear recursive procedure, a computation and space efficient solution to the problem of estimating probabilities from sparse data, and compares favorably to other proposed methods.
Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing
1,264 Citations2013
EMNLP 2010 received 500 submissions, a new record for the conference, and the program committee was able to accept 125 papers in total (an acceptance rate of 25%).
Neural Responding Machine for Short-Text Conversation
1,046 Citations2015Lifeng Shang, Zhengdong Lu +1 more
Empirical study shows that NRM can generate grammatically correct and content-wise appropriate responses to over 75% of the input text, outperforming state-of-the-arts in the same setting, including retrieval-based and SMT-based models.
A Neural Network Approach to Context-Sensitive Generation of Conversational Responses
804 Citations2015Alessandro Sordoni, Michel Galley +7 more
A neural network architecture is used to address sparsity issues that arise when integrating contextual information into classic statistical models, allowing the system to take into account previous dialog utterances.
National Research Council Canada (Government of Canada)Data-Driven Response Generation in Social Media
576 Citations2011Alan Ritter, Colin Cherry +1 more
It is found that mapping conversational stimuli onto responses is more difficult than translating between languages, due to the wider range of possible responses, the larger fraction of unaligned words/phrases, and the presence of large phrase pairs whose alignment cannot be further decomposed.
The MIT Press eBooksLabel Propagation and Quadratic Criterion
281 Citations2006Yoshua Bengio, Olivier Delalleau +1 more
Journal of Artificial Intelligence ResearchA Primer on Neural Network Models for Natural Language Processing
161 Citations2016Yoav Goldberg
This tutorial surveys neural network models from the perspective of natural language processing research, in an attempt to bring natural-language researchers up to speed with the neural techniques.
arXiv (Cornell University)Hierarchical Neural Network Generative Models for Movie Dialogues.
155 Citations2015Iulian Vlad Serban, Alessandro Sordoni +3 more
The recently proposed hierarchical recurrent encoder decoder neural network is extended and demonstrated that it is competitive with state-of-the-art neural language models and backoff n-gram models and can be improved considerably by bootstrapping the learning from a larger questionanswer pair corpus and from pretrained word embeddings.
Predicting Responses to Microblog Posts
101 Citations2012Yoav Artzi, Patrick Pantel +1 more
This work proposes a discriminative model for predicting the likelihood of a response or a retweet on the Twitter network and uses real-world tweets to train models and empirically shows that they are capable of generating accurate predictions for a large number of tweets.
Characterizing and curating conversation threads
97 Citations2013Lars Bäckström, Jon Kleinberg +2 more
A range of approaches are developed and evaluated for two key sub-problems inherent in conversational curation: length prediction and the novel task of re-entry prediction, based on an analysis of the network structure and arrival pattern among the participants, as well as a novel dichotomy in the structure of long threads.
Understanding the Semantic Structure of Noun Phrase Queries
81 Citations2010Xiao Li
This work formally defines the semantic structure of noun phrase queries as comprised of intent heads and intent modifiers and presents methods that automatically identify these constituents as well as their semantic roles based on Markov and semi-Markov conditional random fields.
The MIT Press eBooksLabel Propagation and Quadratic Criterion
74 Citations2006Yoshua Bengio, Delalleau Olivier +1 more
Identifying Intention Posts in Discussion Forums
69 Citations2013Zhiyuan Chen, Bing Liu +3 more
This paper proposes to study the problem of identifying intention posts in online discussion forums and proposes a new transfer learning method which outperforms several strong baselines, including supervised learning in the target domain and a recent transferLearning method.
Hierarchical Label Propagation and Discovery for Machine Generated Email
46 Citations2016James B. Wendt, Michael Bendersky +8 more
This paper formulates an efficient and effective hierarchical label propagation and discovery algorithm that achieves more than 91% precision and 93% recall, while increasing the label coverage by more than 11%.
arXiv (Cornell University)Large Scale Distributed Semi-Supervised Learning Using Streaming\n Approximation
33 Citations2015Sujith Ravi, Qiming Diao
Journal of Web SemanticsDiscovering and understanding word level user intent in Web search queries
25 Citations2014Rishiraj Saha Roy, Rahul Katare +3 more
This paper investigates the association of intent with individual words of a search query and proposes that words in queries can be classified as either content or intent, where content words represent the central topic of the query, while users add intent words to make their requirements more explicit.
arXiv (Cornell University)Large Scale Distributed Semi-Supervised Learning Using Streaming Approximation
10 Citations2015Sujith Ravi, Qiming Diao
This paper presents a novel streaming graph-based SSL approximation that captures the sparsity of the label distribution and ensures the algorithm propagates labels accurately, and further reduces the space complexity per node to $O(1)$.
Revisiting the Predictability of Language: Response Completion in Social Media
9 Citations2012Bo Pang, Sujith Ravi
This work proposes a method for automatic response completion that models both the language used in responses and the specific context provided by the original message, and results show that both components help reduce typing effort.