SMRT Chatbots: Improving Non-Task-Oriented Dialog with Simulated Multiple Reference Training

Abstract

Non-task-oriented dialog models suffer from poor quality and non-diverse responses. To overcome limited conversational data, we apply Simulated Multiple Reference Training (SMRT; Khayrallah et al., 2020), and use a paraphraser to simulate multiple responses per training prompt. We find SMRT improves over a strong Transformer baseline as measured by human and automatic quality scores and lexical diversity. We also find SMRT is comparable to pretraining in human evaluation quality, and outperforms pretraining on automatic quality and lexical diversity, without requiring related-domain dialog data.

EMNLP 2020 Camera Ready

References in corpus (8)

• Distilling the Knowledge in a Neural Network
• Skip-Thought Vectors
• DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset
• Relevance of Unsupervised Metrics in Task-Oriented Dialogue for Evaluating Natural Language Generation
• Generating Informative and Diverse Conversational Responses via Adversarial Information Maximization
• Noisy Parallel Approximate Decoding for Conditional Recurrent Language Model
• Paraphrase Generation as Zero-Shot Multilingual Translation: Disentangling Semantic Similarity from Lexical and Syntactic Diversity
• Paraphrase Augmented Task-Oriented Dialog Generation