Korean Drama Scene Transcript Dataset for Emotion Recognition in Conversations
KD-EmoR Dataset
Korean Drama Scene Transcript Dataset for Emotion Recognition in Conversations
KD-EmoR is socio-behavioral emotion dataset for emotion recognition in realistic conversation scenarios. It consists of total 12289 sentences from 1513 scenes of a Korean TV show named 'Three Brothers'. The dataset is split into Training and testing sets. Each sample consists of sentence_id, person(speaker), sentence, scene_ID, context(Scene description) labeled with one of the following complex emotion labels: euphoria, dysphoria and neutral.
DATASET ACCESS
Access to the dataset is granted through two pathways:
1. Participation in the KERC2022 Challenge
Participation in the KERC2022 challenge automatically grants the permission to the users for using the data in the competition, whereby the participants comply to the usage, distribution and publication policy as stated during the competition.
2. Access Request
To access the dataset and download the files, please send an email to prabesh@jnu.ac.kr, showing a motive to use the dataset for non-commercial academic research purpose. The download link will be provided via email.
PUBLICATION
Sudarshan Pant, Eunchae Lim, Hyung-Jeong Yang, Guee-Sang Lee, Soo-Hyung Kim, Young-Shin Kang, Hyerim Jang, "Korean Drama Scene Transcript Dataset for Emotion Recognition in Conversations," in IEEE Access, vol. 10, pp. 119221-119231, 2022, doi: 10.1109/ACCESS.2022.3221408.
ACKNOWLEDGEMENT
This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT).