Project Conversational Agents
Develop a conversational agent for the cooking domain using DialogFlow, ontology, and natural conversation framework.
Attendance at lab sessions is mandatory. Students will work in groups of 6.
Learning objectives
1. Develop a conversational agent by developing a simple ontology for a knowledge base for the cooking domain which enables the agent to reason about recipes (Applying knowledge and understanding) 2. Apply and understand conversation design patterns (Applying knowledge and understanding) 3. Design and implement visual support for different stages in the conversation (Applying knowledge and understanding) 4. Evaluate a prototype conversational agent for the cooking domain (Making judgments) 5. Plan and organize work together in a project group using appropriate tools (Applying knowledge and understanding)
Students develop a speech-based conversational recipe recommendation agent using MARBEL (Prolog-based agent framework), Python, BERT for NLU, and HTML/Bootstrap for visual support. The project involves building an ASR+NLU pipeline, implementing dialogue management, and creating a visual interface. Groups of 6 students work together over 4 intensive weeks. The exam (15% of grade) is a 1-hour digital exam with 30 multiple choice questions covering the lecture content and preliminaries.
Assessment
The grade is based on the quality of the basic conversational agent (30%), the quality and originality of extensions to the basic agent (30%) and the written report (30%). In addition, students are required to complete an exam that assesses their understanding of course content and their ability to evaluate conversational agents (10%). The exam will be one hour long and proctored by TAs. Each component needs to be at least graded with a 5.5 or higher to pass this course. There are no resit options for the project work, report and exam.
Teaching methods
Lectures, lab sessions (with computer/ laptops). Attendance at labs is obligatory. Students will work in groups of 6.
Literature
Conversational UX Design: A Practitioner's Guide to the Natural Conversation Framework Robert J Moore en Raphael Arar | 29 mei 2019 ISBN-10: 1450363024 ISBN-13: 978-1450363020 A cognitive agent programming guide (online) Learn Prolog Now! (online) Github tutorial (online) Bootstrap tutorial (online) Dialogflow tutorial (online) Natural language understanding models (online)