After following this course, the student can independently:
- describe and discuss current theories and models of dialogue between humans and between humans and artificial agents and analyse observed conversational behaviours in scientific terms
- describe and discuss the state-of-the-art in the automatic analysis and generation of conversational behaviours by (unimodal or multimodal) dialogue systems
- explain the architecture and working of (unimodal or multimodal) dialogue systems and evaluate specific systems
- use tools to annotate verbal and nonverbal conversational behaviours
- apply reliability analysis to annotations of conversational behaviours
- design, implement or simulate and evaluate models of specific conversational behaviours
At the beginning of the course the student is expected to:
Any missing prior knowledge is expected to be acquired through self-study when necessary.
- Be interested in natural language, face-to-face interaction between humans and between humans and machines (artificially intelligent agents)
- Be familiar with basic concepts and methods of natural language processing (NLP), such as part-of-speech classes and part-of-speech tagging, the use of linguistic corpora, grammars and parsing (as explained in the course 201600074 Natural Language Processing) and speech processing such as prosody, speech recognition, speech synthesis (as explained in the course 201600075 Speech Processing). It is not mandatory to have taken these courses.
- Be capable of independently finding relevant literature in addition to the course material.
In this course, we study and discuss different verbal and non-verbal behavioural characteristics, such as speech, gaze and gestures that humans show when communicating face-to-face with both other people and artificial conversational agents. This behaviour is then related to different dialogue functions such as turn-taking, addressing, grounding and backchanneling, that give shape to the communication process and can be implemented in conversational agents to simulate human communication.
The student has to make 2 or 3 homework assignments, at least one of which is an individual assignment (essay) about a specific theme in the theory of conversational interaction, in addition to a small group project where the students work on the development and testing of a simple conversational agent. The topics of lectures and assignments may partially depend on running research projects and challenges. Practical sessions will be provided to learn to work with dialogue annotation tools and to get familiar with conversational agent frameworks that can be used in the project.
The final grade will be determined by the weighted average of the homework assignments and the group project.