Summary
Conversational agents (aka dialogue systems or chatbots) have become one of the most popular and practical applications of AI and natural language processing (NLP): we use them to book tickets over the phone, resolve technical support issues, and answer our questions online. This module provides a thorough introduction to methods for building conversational systems, from rule-based expert systems to neural networks based on large language models; explains the unique challenges that conversational interaction poses for NLP; and looks at key open challenges such as speech, multimodality and real-time interaction.
Aims
Students should expect to learn about the key characteristics of conversational interaction, and how to design, implement and evaluate conversational agents using a range of methods for a range of dialogue tasks.
Syllabus
Introduction to Dialogue Systems
Dialogue Structure & Dialogue Acts
Simple Input/Output Systems
LLMs & Retrieval-Augmented Generation
Dialogue Management
Training End-to-End Systems
(Revision & Feedback Week)
Spoken Dialogue Systems
Clarification & Error-Handling
Real-Time Systems
Multimodal Systems
Agentic Systems & Current Directions
Module Descriptor
https://intranet.eecs.qmul.ac.uk/courses/descriptor/eecsismodule/mod/ECS7033P
Learning Aims and Outcomes
The primary learning outcome in this module is that you will think about learning as a mindset and a process - it has no end point.
By the end of this module you will be able to:
Explain the challenges of conversational interaction for a range of settings
Recommend and explain a design approach for a conversational system for a given setting
Use AI/NLP tools to implement a practical conversational system
Evaluate conversational systems and choose between approaches
Conversational agents (aka dialogue systems or chatbots) have become one of the most popular and practical applications of AI and natural language processing (NLP): we use them to book tickets over the phone, resolve technical support issues, and answer our questions online. This module provides a thorough introduction to methods for building conversational systems, from rule-based expert systems to neural networks based on large language models; explains the unique challenges that conversational interaction poses for NLP; and looks at key open challenges such as speech, multimodality and real-time interaction.
Aims
Students should expect to learn about the key characteristics of conversational interaction, and how to design, implement and evaluate conversational agents using a range of methods for a range of dialogue tasks.
Syllabus
Introduction to Dialogue Systems
Dialogue Structure & Dialogue Acts
Simple Input/Output Systems
LLMs & Retrieval-Augmented Generation
Dialogue Management
Training End-to-End Systems
(Revision & Feedback Week)
Spoken Dialogue Systems
Clarification & Error-Handling
Real-Time Systems
Multimodal Systems
Agentic Systems & Current Directions
Module Descriptor
https://intranet.eecs.qmul.ac.uk/courses/descriptor/eecsismodule/mod/ECS7033P
Learning Aims and Outcomes
The primary learning outcome in this module is that you will think about learning as a mindset and a process - it has no end point.
By the end of this module you will be able to:
Explain the challenges of conversational interaction for a range of settings
Recommend and explain a design approach for a conversational system for a given setting
Use AI/NLP tools to implement a practical conversational system
Evaluate conversational systems and choose between approaches
