A Simple, Fast Diverse Decoding Algorithm for Neural Generation

Abstract

In this paper, we propose a simple, fast decoding algorithm that fosters diversity in neural generation. The algorithm modifies the standard beam search algorithm by adding an inter-sibling ranking penalty, favoring choosing hypotheses from diverse parents. We evaluate the proposed model on the tasks of dialogue response generation, abstractive summarization and machine translation. We find that diverse decoding helps across all tasks, especially those for which reranking is needed. We further propose a variation that is capable of automatically adjusting its diversity decoding rates for different inputs using reinforcement learning (RL). We observe a further performance boost from this RL technique. This paper includes material from the unpublished script "Mutual Information and Diverse Decoding Improve Neural Machine Translation" (Li and Jurafsky, 2016).

References in corpus (13)

• Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
• Microsoft COCO Captions: Data Collection and Evaluation Server
• A Neural Conversational Model
• Sequence Level Training with Recurrent Neural Networks
• How NOT To Evaluate Your Dialogue System: An Empirical Study of Unsupervised Evaluation Metrics for Dialogue Response Generation
• On Using Monolingual Corpora in Neural Machine Translation
• A Hierarchical Neural Autoencoder for Paragraphs and Documents
• Reinforcement Learning Neural Turing Machines - Revised
• A Neural Network Approach to Context-Sensitive Generation of Conversational Responses
• LSTM based Conversation Models
• Noisy Parallel Approximate Decoding for Conditional Recurrent Language Model
• Learning to Compose Words into Sentences with Reinforcement Learning
• Agreement-based Joint Training for Bidirectional Attention-based Neural Machine Translation

Cited by in corpus (75)

• The Curious Case of Neural Text Degeneration
• Wizard of Wikipedia: Knowledge-Powered Conversational agents
• A Survey on Dialogue Systems: Recent Advances and New Frontiers
• TransferTransfo: A Transfer Learning Approach for Neural Network Based Conversational Agents
• Neural Text Generation with Unlikelihood Training
• Adversarial Learning for Neural Dialogue Generation
• Emotional Chatting Machine: Emotional Conversation Generation with Internal and External Memory
• Fast Abstractive Summarization with Reinforce-Selected Sentence Rewriting
• Neural Abstractive Text Summarization with Sequence-to-Sequence Models
• Learning to Decode for Future Success
• Challenges in Building Intelligent Open-domain Dialog Systems
• Data Diversification: A Simple Strategy For Neural Machine Translation
• DialogVED: A Pre-trained Latent Variable Encoder-Decoder Model for Dialog Response Generation
• MetaMT,a MetaLearning Method Leveraging Multiple Domain Data for Low Resource Machine Translation
• Neural Text Generation: A Practical Guide
• An Ensemble Model with Attention Based Mechanism for Image Captioning
• Task-Oriented Dialog Systems that Consider Multiple Appropriate Responses under the Same Context
• Neural Language Generation: Formulation, Methods, and Evaluation
• Stochastic Beams and Where to Find Them: The Gumbel-Top-k Trick for Sampling Sequences Without Replacement
• A Condense-then-Select Strategy for Text Summarization
• Trainable Greedy Decoding for Neural Machine Translation
• Sequence to Sequence Mixture Model for Diverse Machine Translation
• Mirostat: A Neural Text Decoding Algorithm that Directly Controls Perplexity
• Controlling Decoding for More Abstractive Summaries with Copy-Based Networks
• Addressing Semantic Drift in Question Generation for Semi-Supervised Question Answering
• The Garden of Forking Paths: Towards Multi-Future Trajectory Prediction
• Non-Autoregressive Neural Dialogue Generation
• Deep Active Learning for Dialogue Generation
• LAVA NAT: A Non-Autoregressive Translation Model with Look-Around Decoding and Vocabulary Attention
• Accelerating Real-Time Question Answering via Question Generation
• Diverse Beam Search for Increased Novelty in Abstractive Summarization
• Later-stage Minimum Bayes-Risk Decoding for Neural Machine Translation
• Learning to Coordinate Multiple Reinforcement Learning Agents for Diverse Query Reformulation
• Large-scale Pretraining for Neural Machine Translation with Tens of Billions of Sentence Pairs
• Generating Multiple Diverse Responses for Short-Text Conversation
• Generating Multiple Diverse Responses with Multi-Mapping and Posterior Mapping Selection
• Mixture Content Selection for Diverse Sequence Generation
• Conditional Text Generation for Harmonious Human-Machine Interaction
• Complexity-Weighted Loss and Diverse Reranking for Sentence Simplification
• Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints
• Why are Sequence-to-Sequence Models So Dull? Understanding the Low-Diversity Problem of Chatbots
• Better Conversations by Modeling,Filtering,and Optimizing for Coherence and Diversity
• A Stable and Effective Learning Strategy for Trainable Greedy Decoding
• Teaching Machines to Converse
• Why Do Neural Response Generation Models Prefer Universal Replies?
• Affective Neural Response Generation
• Variational Prefix Tuning for Diverse and Accurate Code Summarization Using Pre-trained Language Models
• Diverse, Controllable, and Keyphrase-Aware: A Corpus and Method for News Multi-Headline Generation
• Sparse Text Generation
• Self-Training Sampling with Monolingual Data Uncertainty for Neural Machine Translation
• Target Guided Emotion Aware Chat Machine
• Improving the Robustness of QA Models to Challenge Sets with Variational Question-Answer Pair Generation
• Diversifying Dialogue Generation with Non-Conversational Text
• A Discrete CVAE for Response Generation on Short-Text Conversation
• Lost in Machine Translation: A Method to Reduce Meaning Loss
• From Recognition to Prediction: Analysis of Human Action and Trajectory Prediction in Video
• Discrete Structural Planning for Neural Machine Translation
• Cue-word Driven Neural Response Generation with a Shrinking Vocabulary
• Paraphrase Generation as Unsupervised Machine Translation
• Generating Diverse Translation by Manipulating Multi-Head Attention
• HSCJN: A Holistic Semantic Constraint Joint Network for Diverse Response Generation
• Keyword-Guided Neural Conversational Model
• Toward Improving Coherence and Diversity of Slogan Generation
• Prediction, Selection, and Generation: Exploration of Knowledge-Driven Conversation System
• Multi-Turn Beam Search for Neural Dialogue Modeling
• Towards Efficiently Diversifying Dialogue Generation via Embedding Augmentation
• multiPRover: Generating Multiple Proofs for Improved Interpretability in Rule Reasoning
• ConRPG: Paraphrase Generation using Contexts as Regularizer
• Domain-Constrained Advertising Keyword Generation
• REAM$♯$: An Enhancement Approach to Reference-based Evaluation Metrics for Open-domain Dialog Generation
• Generating Informative Dialogue Responses with Keywords-Guided Networks
• Single-Queue Decoding for Neural Machine Translation
• Predict and Use Latent Patterns for Short-Text Conversation
• Generating Diverse Translation from Model Distribution with Dropout
• Grounding Conversations with Improvised Dialogues