Corpus ID: 237504588

Controllable Dialogue Generation with Disentangled Multi-grained Style Specification and Attribute Consistency Reward

@article{Hu2021ControllableDG,
  title={Controllable Dialogue Generation with Disentangled Multi-grained Style Specification and Attribute Consistency Reward},
  author={Zhe Hu and Zhiwei Cao and Hou Pong Chan and Jiachen Liu and Xinyan Xiao and Jinsong Su and Hua Wu},
  journal={ArXiv},
  year={2021},
  volume={abs/2109.06717}
}
  • Zhe Hu, Zhiwei Cao, +4 authors Hua Wu
  • Published 2021
  • Computer Science
  • ArXiv
Controllable text generation is an appealing but challenging task, which allows users to specify particular attributes of the generated outputs. In this paper, we propose a controllable dialogue generation model to steer response generation under multi-attribute constraints. Specifically, we define and categorize the commonly-used control attributes into global and local ones, which possess different granularities of effects on response generation. Then, we significantly extend the conventional… Expand

References

SHOWING 1-10 OF 53 REFERENCES
Sentence-Level Content Planning and Style Specification for Neural Text Generation
TLDR
This work presents an end-to-end trained two-step generation model, where a sentence-level content planner first decides on the keyphrases to cover as well as a desired language style, followed by a surface realization decoder that generates relevant and coherent text. Expand
Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints
TLDR
This work proposes a simple yet effective approach for incorporating side information in the form of distributional constraints over the generated responses that generates responses that are much less generic without sacrificing plausibility. Expand
PLATO: Pre-trained Dialogue Generation Model with Discrete Latent Variable
TLDR
This work proposes a novel dialogue generation pre-training framework to support various kinds of conversations, including chit-chat, knowledge grounded dialogues, and conversational question answering, and introduces discrete latent variables to tackle the inherent one-to-many mapping problem in response generation. Expand
Structuring Latent Spaces for Stylized Response Generation
TLDR
StyleFusion is proposed, which bridges conversation modeling and non-parallel style transfer by sharing a structured latent space that allows the system to generate stylized relevant responses by sampling in the neighborhood of the conversation model prediction, and continuously control the style level. Expand
Learning to Control the Specificity in Neural Response Generation
TLDR
A novel controlled response generation mechanism to handle different utterance-response relationships in terms of specificity is proposed, which introduces an explicit specificity control variable into a sequence-to-sequence model, which interacts with the usage representation of words through a Gaussian Kernel layer. Expand
Steering Output Style and Topic in Neural Response Generation
TLDR
The decompose the neural generation process into empirically easier sub-problems: a faithfulness model and a decoding method based on selective-sampling and training and sampling algorithms that bias the generation process with a specific language style restriction, or a topic restriction. Expand
Neural Response Generation with Meta-words
TLDR
A novel goal-tracking memory network is proposed that formalizes meta-word expression as a goal in response generation and manages the generation process to achieve the goal with a state memory panel and a state controller. Expand
Deep Reinforcement Learning For Modeling Chit-Chat Dialog With Discrete Attributes
TLDR
This paper proposes to formulate the dialog attribute prediction as a reinforcement learning (RL) problem and use policy gradients methods to optimize utterance generation using long-term rewards to make the policy optimization more practical and sample efficient. Expand
CTRL: A Conditional Transformer Language Model for Controllable Generation
TLDR
CTRL is released, a 1.63 billion-parameter conditional transformer language model, trained to condition on control codes that govern style, content, and task-specific behavior, providing more explicit control over text generation. Expand
Generating Multiple Diverse Responses for Short-Text Conversation
TLDR
A novel response generation model is proposed, which considers a set of responses jointly and generates multiple diverse responses simultaneously simultaneously and a reinforcement learning algorithm is designed to solve the model. Expand
...
1
2
3
4
5
...