End-to-End Trainable Non-Collaborative Dialog System

• DOI: 10.48550/arxiv.1911.10742
• Publication Date: 2019-01-01
• Authors: Yu Li; Kun Qian; Weiyan Shi; Zhou Yu
• Keywords: Representation

End-to-end task-oriented dialog models have achieved promising performance on collaborative tasks where users willingly coordinate with the system to complete a given task. While in non-collaborative settings, for example, negotiation and persuasion, systems do not share common goal. As result, compared collaborate tasks, people use social content build rapport trust these settings order advance their goals. To handle content, we introduce hierarchical intent annotation scheme, which can be generalized different tasks. Building upon TransferTransfo (Wolf et al. 2019), propose an end-to-end neural network model generate diverse coherent responses. Our utilizes semantic slots as intermediate sentence representation guide generation process. In addition, design filter select appropriate responses based whether representations fit designed task conversation constraints. guides while simultaneously keeps them engaged. We test our approach newly proposed ANTISCAM dataset existing PERSUASIONFORGOOD dataset. Both automatic human evaluations suggest that outperforms multiple baselines two