Service providers are increasingly exploring the use of conversational agent (CA) or dialog-based system to support end customers, as a CA promises natural method for users to interact and a convenient channel for customer service.
Commercial CAs excel in addressing narrow domain-specific tasks that entail a limited number of user interactions such as searching for restaurants, providing location directions, or scheduling meetings.
Designing a CA that can support customers with queries on services of a large service system, such as a banking or an insurance system, requires sufficient knowledge of its services such as the service capabilities, constraints, and business service rules in addition to understanding user utterances. The design of a CA is typically an independent activity and its linkages to the service system it supports are left to the designers.
A widely successful dialog design approach known as frame-based dialog system is based on the design of frames (or entities) and slots (attribute, and values) for a user intent or goal. The design of user intents, frames and slots can be challenging for large service systems.
The actions that a CA needs to take to support user requests may need information from multiple knowledge sources and software systems involved in the service. The CA action design has had little attention for such systems.
Designing a CA requires linkages to business process, rules and software systems. The design would need to be informed by learning intents, slots from large corpus user-user conversations or user-system conversations.
Research activities will vary depending on your interests, skills, and the duration of the project.
The research activities include the following:
- a literature review on CA and design of user intents and fulfillment approaches that use existing dialog corpus (VRES)
- an investigation of methods that are based on the corpus of human conversations, knowledge bases, service design or an ensemble of multiple methods (VRES, Honours)
- the development of intent and action design for CA (Masters)
- the development, implementation, strategies and programs that will provide capabilities to design CAs based on a service system accommodating a combination of methods. (PhD)
The expected project outcomes are dependent on the scope of the project that you undertake.
Key outcomes include the following:
- gap analysis in the design of CAs for large service systems.
- evaluation of existing CA design techniques that are based on learning from a corpus of data, or domain based knowledge models.
- new techniques, methods for designing a CA for a service system.
Skills and experience
For this project we expect you to have the following skills and experiences:
- familiarity within the fields of software or service design
- some familiarity with machine learning
- reasonable writing skills
- programming skills, preferably in Python.
You may be able to apply for a research scholarship in our annual scholarship round.
Contact the supervisor for more information.