Learning from Implicit User Feedback, Emotions and Demographic Information in Task-Oriented and Document-Grounded Dialogues
This work addresses the need for better dialogue systems in specific domains like task-oriented and document-grounded settings, but it is incremental as it builds on existing methods by introducing a new dataset.
The authors tackled the problem of improving task completion and factual consistency in task-oriented and document-grounded dialogues by incorporating implicit user feedback, emotions, and demographic information, resulting in positive impacts such as more informative, relevant, and factually consistent responses as reported in human evaluations.
Implicit user feedback, user emotions and demographic information have shown to be promising sources for improving the accuracy and user engagement of responses generated by dialogue systems. However, the influence of such information on task completion and factual consistency, which are important criteria for task-oriented and document-grounded dialogues, is not yet known. To address this, we introduce FEDI, the first English task-oriented and document-grounded dialogue dataset annotated with this information. Our experiments with Flan-T5, GPT-2 and Llama 2 show a particularly positive impact on task completion and factual consistency. Participants in our human evaluation reported that the responses generated by the feedback-trained models were more informative (Flan-T5 and GPT-2), relevant and factual consistent (Llama 2).