Deadline: July 31st
Conversational AI interfaces are beginning to allow people to communicate naturally with computers for the first time. A new generation of conversational search is emerging with the rise of virtual personal assistants (Google Assistant, Alexa, Siri, Cortana, and others) using voice-based search interfaces. However, currently, these systems are only useful in narrow domains and support limited information tasks. In contrast, our objective is to research approaches that will enable conversational information seeking (CIS) agent systems to assist users accomplishing (complex) task-oriented information goals across all domains of human knowledge. The research has a diverse set of applications areas including: health, shopping, customer service, technical support, travel planning, and many others. The goal is to transform how humans and algorithms engage with information and each other.
The topic of this PhD is focused on modeling and improving the task-oriented sequence of exchanges between one or more users and a conversational information seeking agent. There are many open challenges across diverse areas: learning information access policies with deep reinforcement learning (RL), personal knowledge graph construction, new models of interactive feedback and disambiguation, cross-document information extraction and summarization, explainable and transparent models for agent trust, conversational information seeking user modelling, and others. The target research area will depend on the candidate’s experience and interests.
The successful candidate will design, develop, and evaluate conversational information seeking agent systems. The candidate is expected to use and develop new novel deep learning models for subtasks and end-to-end systems.
Email me at Jeff.Dalton [AT] glasgow.ac.uk for further information and to apply.