Dialogue Management and Language Generation for a Robust Conversational Virtual Coach: Validation and User Study.

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence
• Computational Linguistics

Background:

• Designing interactive systems requires interdisciplinary collaboration.
• Voice-based Spoken Dialogue Systems (SDS) are typically used for simpler tasks.
• Coaching conversations present unique challenges for SDS due to complexity.

Purpose of the Study:

• To develop core modules for a Voice-based Spoken Dialogue System (SDS).
• To create an intelligent and robust Virtual Coach for complex coaching conversations.
• To integrate these modules into a functional and effective system.

Main Methods:

• Development of a Dialogue Manager and a Language Generator.
• Implementation of an efficient system integration procedure.
• User study validation with elderly participants in multiple countries.

Main Results:

• The developed system successfully functions as a Virtual Coach.
• The system handles challenging, long, and complex coaching conversations.
• Successful validation across three countries (Spain, France, Norway) with diverse users.

Conclusions:

• The proposed SDS architecture is effective for advanced coaching applications.
• The Virtual Coach demonstrates robustness and adaptability across different languages and cultures.
• This work advances the capability of SDS in specialized, complex interaction domains.