Modeling Mentor-Mentee Dialogues in Film

Area of Science:

• Artificial Intelligence
• Computational Linguistics
• Film Studies

Background:

• Designing synthetic agents for coherent, in-character conversations requires understanding human dialogue dynamics.
• Previous work focused on mentor strategies; this study expands to include mentee roles for a comprehensive model.

Purpose of the Study:

• To inform the design of a synthetic agent capable of consistent, mentor-like conversations with humans.
• To develop an extended model accounting for both mentor and mentee conversational activities.
• To formalize joint communication actions and goals between characters.

Main Methods:

• Data-driven analysis of verbal communication between fictional mentor and mentee characters in films.
• Introduction of categories for intents, projects, and relationship phases to formalize communication.
• Qualitative and quantitative analysis of an annotated corpus of character utterances.
• Evaluation of state-of-the-art automated utterance classification approaches.

Main Results:

• An extended model was developed, incorporating both mentor and mentee dialogue contributions.
• Categories of intents, projects, and relationship phases were defined to analyze joint communication.
• The annotated corpus provides a foundation for understanding and replicating in-character dialogue.

Conclusions:

• Modeling both sides of a dialogue is crucial for creating realistic synthetic agents.
• The developed framework and corpus support advancements in automated in-character dialogue generation.
• This research contributes to more natural and engaging human-AI conversational experiences.