The talk "User satisfaction/ Reward Function" by Hsien-Chin Lin takes place on 14.05.2021 at 14:00 h as a part of the seminar “Selected topics in Machine Learning and Natural Language Processing". Please email to enrol and get further details.
Please, click here for online participation.
Users are central to dialogue systems development because these systems need to take into account the users’ behaviour when choosing actions. The users act differently depending on whether they are content or discontent with the system. This means that the dialogue system needs to optimise it policy based on not only task success but also on user satisfaction. In this talk, we will look at different methods to model user satisfaction, how to use them to improve the dialogue system and the challenges of using user satisfaction as a reward function for dialogue policy optimisation.