ValueNet: A New Dataset for Human Value Driven Dialogue System
Liang Qiu, Yizhou Zhao, Jinchao Li, Pan Lu, Baolin Peng, Jianfeng Gao,, Song-Chun Zhu

TL;DR
ValueNet introduces a large-scale dataset of human attitudes on scenarios based on human values, enabling the development of value-driven dialogue systems that enhance empathy and personalization.
Contribution
It is the first large-scale dataset for human value modeling and integrates a value model into emotionally intelligent dialogue systems.
Findings
The value model improves personalized dialogue generation.
Reinforcement learning with value rewards achieves state-of-the-art results.
Values as features enhance emotion recognition and empathetic responses.
Abstract
Building a socially intelligent agent involves many challenges, one of which is to teach the agent to speak guided by its value like a human. However, value-driven chatbots are still understudied in the area of dialogue systems. Most existing datasets focus on commonsense reasoning or social norm modeling. In this work, we present a new large-scale human value dataset called ValueNet, which contains human attitudes on 21,374 text scenarios. The dataset is organized in ten dimensions that conform to the basic human value theory in intercultural research. We further develop a Transformer-based value regression model on ValueNet to learn the utility distribution. Comprehensive empirical results show that the learned value model could benefit a wide range of dialogue tasks. For example, by teaching a generative agent with reinforcement learning and the rewards from the value model, our…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
Taxonomy
TopicsAI in Service Interactions · Topic Modeling · Persona Design and Applications
