PicPersona-TOD: A Dataset for Personalizing Utterance Style in Task-Oriented Dialogue with Image Persona
This is the official code for PicPersona-TOD, accepted in NAACL 2025 Main
Jihyun Lee, Yejin Jeon, Seungyeon Seo, Gary Geunbae Lee
PicPersona-TOD is a dataset designed for generating personalized responses in task-oriented dialogue using image-based persona information.
Final dataset is in
/dataset/mwoz_sgd
TBD
For questions, feel free to reach out to jihyunlee@postech.ac.kr.