Links:

Unit 1 Introducing Dialogue Systems

Unit 2 Rule-based Dialogue Systems

Unit 3 Statistical Data-driven Dialogue Systems

Unit 4 Evaluating Dialogue Systems

Unit 5 End-to-End Neural Dialogue Systems

Syllabus

UNIT - I: Introducing Dialogue Systems

• Foundations: Introduction to Dialogue Systems and their evolution.

• History: The lineage of Conversational AI from early iterations to modern tech.

• State of the Art: Overview of Present-Day Dialogue Systems.

• Modeling & Design: * Conceptualizing Conversation Dialogue Systems.

  • Best practices for Designing and Developing Dialogue Systems.

UNIT - II: Rule-Based Dialogue Systems: Architecture, Methods, and Tools

• System Design: Understanding Dialogue Systems Architecture.

• Development Workflow: Designing a Dialogue System from the ground up.

• Tech Stack: Tools and frameworks for developing Rule-Based systems.

• Case Study: Rule-Based Techniques in Dialogue Systems participating in the Alexa Prize.

UNIT - III: Statistical Data-Driven Dialogue Systems

• Approach: Motivating the shift toward Statistical Data-Driven models.

• Core Components: Dialogue components within the statistical framework.

• Decision Processes:

  • Reinforcement Learning (RL) fundamentals.

  • Representing Dialogue as a Markov Decision Process (MDP).

  • Transitioning from MDPs to POMDPs (Partially Observable MDPs).

• Management: * Dialogue State Tracking (DST).

  • Dialogue Policy optimization.

  • Challenges and issues with RL in POMDP environments.

UNIT - IV: Evaluating Dialogue Systems

• Evaluation Metrics: The process and methodology of evaluation.

• System Types:

  • Evaluating Task-Oriented Dialogue Systems.

  • Evaluating Open-Domain Dialogue Systems.

• Frameworks: * PARADISE Framework.

  • Quality of Experience (QoE) and Interaction Quality.

• Synthesis: Determining the best practices for comprehensive system evaluation.

UNIT - V: End-to-End Neural Dialogue Systems

• Neural Modeling: Introduction to Neural Network approaches in dialogue.

• Architectures:

  • Neural Conversational Models.

  • Retrieval-Based vs. Generative Response Generation.

• Applications:

  • Task-Oriented Neural Dialogue Systems.

  • Open-Domain Neural Dialogue Systems (Chatbots).

• Current Landscape: * Contemporary issues and existing solutions.

  • Datasets, Competitions, Tasks, and Challenges in the field.

Based on:

Michael McTear, “Conversational AI: Dialogue Systems, Conversational Agents, and Chatbots”, Second Edition, Moran and Claypool Publishers, 2020.

and

Google Gemini’s Summarization