Learning dialogue strategies for interactive database search
Speaker:Verena Rieser
Institution: Saarland University
Abstract:
The topic of my PhD thesis to optimise Multimodal Clarification Strategies for Information Seeking Dialogue Systems using Reinforcement Learning. In this talk a "prove of concept" is given that dialogue strategies for a wide spectrum of applications can be learnt that are significantly better than hand-coded rules.
We show how to learn optimal dialogue policies for a wide range of database search applications, concerning how many database search results to present to the user, and when to present them. We use Reinforcement Learning methods for a wide spectrum of different database simulations, turn penalty conditions, and noise conditions. Our objective is to show that our policy learning framework covers this spectrum. We can show that even for challenging cases learning significantly outperforms hand-coded policies tailored to the different operating situations. The polices are adaptive/context-sensitive in respect of both the overall operating situation (e.g. noise) and the local context of the interaction (e.g. user's last move). The learned policies produce an average relative increase in reward of 25.7% over the corresponding threshold-based hand-coded baseline policies.
In current work this general framework will be applied to one specific application, namely a dialogue for music search in the SAMMIE domain. If time allows I will report on recent results.