DOI QR코드

DOI QR Code

Detects depression-related emotions in user input sentences

사용자 입력 문장에서 우울 관련 감정 탐지

  • Oh, Jaedong (Artificial Inteligence Convergence, Sungkyunkwan University) ;
  • Oh, Hayoung (College of Computing and Informatics, Sungkyunkwan University)
  • Received : 2022.09.19
  • Accepted : 2022.12.04
  • Published : 2022.12.31

Abstract

This paper proposes a model to detect depression-related emotions in a user's speech using wellness dialogue scripts provided by AI Hub, topic-specific daily conversation datasets, and chatbot datasets published on Github. There are 18 emotions, including depression and lethargy, in depression-related emotions, and emotion classification tasks are performed using KoBERT and KOELECTRA models that show high performance in language models. For model-specific performance comparisons, we build diverse datasets and compare classification results while adjusting batch sizes and learning rates for models that perform well. Furthermore, a person performs a multi-classification task by selecting all labels whose output values are higher than a specific threshold as the correct answer, in order to reflect feeling multiple emotions at the same time. The model with the best performance derived through this process is called the Depression model, and the model is then used to classify depression-related emotions for user utterances.

본 논문은 AI Hub에서 제공하는 웰니스 대화 스크립트, 주제별 일상 대화 데이터세트와 Github에 공개된 챗봇 데이터세트를 활용하여 사용자의 발화에서 우울 관련 감정을 탐지하는 모델을 제안한다. 우울 관련 감정에는 우울감, 무기력을 비롯한 18가지 감정이 존재하며, 언어 모델에서 높은 성능을 보이는 KoBERT와 KoELECTRA 모델을 사용하여 감정 분류 작업을 수행한다. 모델별 성능 비교를 위해 우리는 데이터세트를 다양하게 구축하고, 좋은 성능을 보이는 모델에 대해 배치 크기와 학습률을 조정하면서 분류 결과를 비교한다. 더 나아가, 사람은 동시에 여러 감정을 느끼는 것을 반영하기 위해, 모델의 출력값이 특정 임계치보다 높은 레이블들을 모두 정답으로 선정함으로써, 다중 분류 작업을 수행한다. 이러한 과정을 통해 도출한 성능이 가장 좋은 모델을 Depression model이라 부르며, 이후 사용자 발화에 대해 우울 관련 감정을 분류할 때 해당 모델을 사용한다.

Keywords

Acknowledgement

This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. NRF-2022R1F1A1074696).

References

  1. A red light for modern mental health and stress management are essential [Internet]. Available:http://www.medical-tribune.co.kr/news/articleView.html?idxno=100431.
  2. J. Q. Yuki, Md. M. Q. Sakib, Z. Zamal, S. H. Efel, and M. A. Khan, "Detecting Depression from Human Conversations," in Proceedings of the 8th International Conference on Computer and Communications Management (ICCCM'20), New York: NY, USA, pp. 14-18, 2020.
  3. A story about our brains [Internet]. Available: http://scienceon.hani.co.kr/436471.
  4. Wellness conversation dataset [Internet]. Available: https://aihub.or.kr/opendata/keti-data/recognition-laguage/KETI-02-006.
  5. Subject-specific text Daily conversation [Internet]. Available: https://aihub.or.kr/aihubdata/data/view.do?currMenu=115&topMenu=100&aihubDataSe=realm&dataSetSn=543.
  6. Korean Chatbot dataset [Internet]. Available: https://github.com/songys/Chatbot_data.
  7. H. Chin, G. Baek, C. Cha, J. Choi, H. Im, and M. Cha, "A study on the categories and characteristics of depressive moods in chatbot data," in Proceedings of the Korea Information Processing Society Conference, Yeosu, Korea, pp. 993-996, 2021.
  8. S. J. Sohn, M. S. Park, J. E. Park, and J. H. Sohn, "Korean Emotion Vocabulary: Extraction and Categorization of Feeling Words," Science of Emotion and Sensibility, vol. 15, no. 1, pp. 105-120, Mar. 2012.
  9. S. H. An and O. R. Jeong "A Study on the Psychological Counseling AI Chatbot System based on Sentiment Analysis," Journal of Information Technology Services, vol. 20, no. 3, pp. 75-86, Jun. 2021. https://doi.org/10.9716/KITS.2021.20.3.075
  10. Y. G. Song, K. M. Jung, and H. Lee, "A BERGPT-chatbot for mitigating negative emotions," Journal of the Korea Society of Computer and Information, vol. 26, no. 12, pp. 53-59, Dec. 2021. https://doi.org/10.9708/JKSCI.2021.26.12.053
  11. A. Sharma, A. S. Miner, D. C. Atkins, and T. Althoff, "A Computational Approach to Understanding Empathy Expressed in Text-Based Mental Health Support," in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), Online, pp. 5263-5276, 2020.
  12. J. Devlin, M. W. Chang, K. Lee, and K. Toutanova, "Bert: Pre-training of Deep Bidirectional Transformers for Language Understanding," in Proceedings of NAACL-HLT 2019, Minneapolis: MN, USA, pp. 4171-4186, 2018.
  13. About BERT [Internet]. Available: https://happy-obok.tistory.com/23.
  14. K. Clark, M. T. Luong, Q. V. Le, and C. D. Manning, "ELECTRA: Pre-training text encoders as discriminators rather than generators," arXiv preprint arXiv:2003.10555, Mar. 2020.
  15. I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, "Generative adversarial networks," Communications of the ACM, vol. 63, no. 11, pp. 139-144, Oct. 2020. https://doi.org/10.1145/3422622
  16. I. Kandel and M. Castelli, "The effect of batch size on the generalizability of the convolutional neural networks on a histopathology dataset," ICT Express, vol. 6, no. 4, pp. 312-315, Dec. 2020. https://doi.org/10.1016/j.icte.2020.04.010