TY - GEN
T1 - Tell Me More, Tell Me More
T2 - 32nd IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2023
AU - Chierici, Alberto
AU - Habash, Nizar
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Time-Offset Interaction Applications (TOIAs) are narrative-sharing systems that use databases of previously recorded videos of real people to mimic conversations with them. These video databases comprise large (the larger, the better) collections of videos of answers paired with specific questions. This paper focuses on a solution to the challenge of creating such databases without exhausting their creators' creativity, energy, and interest. We describe the design and development process of Question Suggester (QS)-an intelligent GPT-3-based service that generates suggested questions following up a conversation based on the history of recorded questions and answers. We conduct a user study to empirically evaluate the value of QS for reducing the effort to create a video database while creating an interaction that is enjoyable. The users' average experience rating for QS is 4.6 compared to 4.0 when QS is not used (on a 1-5 scale, p-value<0.05). The experience with interactions so created is more enjoyable, too (3.7vs. 3.3, p-value<0.05). The usage metrics and qualitative feedback confirm that QS is essential for interactive video-recording systems and for increasing their adoption.
AB - Time-Offset Interaction Applications (TOIAs) are narrative-sharing systems that use databases of previously recorded videos of real people to mimic conversations with them. These video databases comprise large (the larger, the better) collections of videos of answers paired with specific questions. This paper focuses on a solution to the challenge of creating such databases without exhausting their creators' creativity, energy, and interest. We describe the design and development process of Question Suggester (QS)-an intelligent GPT-3-based service that generates suggested questions following up a conversation based on the history of recorded questions and answers. We conduct a user study to empirically evaluate the value of QS for reducing the effort to create a video database while creating an interaction that is enjoyable. The users' average experience rating for QS is 4.6 compared to 4.0 when QS is not used (on a 1-5 scale, p-value<0.05). The experience with interactions so created is more enjoyable, too (3.7vs. 3.3, p-value<0.05). The usage metrics and qualitative feedback confirm that QS is essential for interactive video-recording systems and for increasing their adoption.
UR - http://www.scopus.com/inward/record.url?scp=85186989271&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85186989271&partnerID=8YFLogxK
U2 - 10.1109/RO-MAN57019.2023.10309424
DO - 10.1109/RO-MAN57019.2023.10309424
M3 - Conference contribution
AN - SCOPUS:85186989271
T3 - IEEE International Workshop on Robot and Human Communication, RO-MAN
SP - 1725
EP - 1730
BT - 2023 32nd IEEE International Conference on Robot and Human Interactive Communication, RO-MAN 2023
PB - IEEE Computer Society
Y2 - 28 August 2023 through 31 August 2023
ER -