The Journal of the Convergence on Culture Technology (문화기술의 융합)

The International Promotion Agency of Culture Technology (국제문화기술진흥원)
권호 표지
DOI QR코드

DOI QR Code

A Study on the Development of Emotional Content through Natural Language Processing Deep Learning Model Emotion Analysis

자연어 처리 딥러닝 모델 감정분석을 통한 감성 콘텐츠 개발 연구

  • 이현수 (성결대학교 미디어소프트웨어학과) ;
  • 김민하 (성결대학교 미디어소프트웨어학과) ;
  • 서지원 (성결대학교 미디어소프트웨어학과) ;
  • 김정이 (성결대학교 미디어소프트웨어학과)
  • Received : 2023.06.25
  • Accepted : 2023.07.10
  • Published : 2023.07.31
Download PDF
⟨ Previous Next ⟩

Abstract

We analyze the accuracy of emotion analysis of natural language processing deep learning model and propose to use it for emotional content development. After looking at the outline of the GPT-3 model, about 6,000 pieces of dialogue data provided by Aihub were input to 9 emotion categories: 'joy', 'sadness', 'fear', 'anger', 'disgust', and 'surprise'. ', 'interest', 'boredom', and 'pain'. Performance evaluation was conducted using the evaluation indices of accuracy, precision, recall, and F1-score, which are evaluation methods for natural language processing models. As a result of the emotion analysis, the accuracy was over 91%, and in the case of precision, 'fear' and 'pain' showed low values. In the case of reproducibility, a low value was shown in negative emotions, and in the case of 'disgust' in particular, an error appeared due to the lack of data. In the case of previous studies, emotion analysis was mainly used only for polarity analysis divided into positive, negative, and neutral, and there was a limitation in that it was used only in the feedback stage due to its nature. We expand emotion analysis into 9 categories and suggest its use in the development of emotional content considering it from the planning stage. It is expected that more accurate results can be obtained if emotion analysis is performed by additionally collecting more diverse daily conversations through follow-up research.

본 연구는 자연어 처리 딥러닝 모델의 감정분석 정확성을 확인해보고 이를 감성 콘텐츠 개발에 활용하도록 제안한다. GPT-3모델의 개요를 살펴본 후 Aihub에서 제공하는 희곡 대사 데이터 약 6000개를 입력하고 '기쁨', '슬픔', '공포', '분노', '혐오', '놀람', '흥미', '지루함', '통증' 총 9가지 감정 범주로 분류하였다. 이후 자연어 처리 모델 평가 방법인 정확도, 정밀도, 재현율, F1-score 의 평가지표를 활용하여 성능평가를 진행하였다. 감정분석 결과 91% 이상의 정확도를 보였으며 정밀도의 경우 '공포','통증'이 낮은 수치를 보였다. 재현도의 경우 '슬픔', '분노', '혐오'와 같은 부정적인 감정에서 낮은 수치가 나타났고 특히 '혐오'의 경우 데이터 양의 부족으로 인해 오차가 나타난 것으로 확인된다. 기존 연구의 경우 감정분석을 긍정, 부정, 중립으로 나누는 극성분석에만 주로 사용되어 그 특성상 피드백 단계에서만 사용되는 한계가 있었다. 본 연구는 감정분석을 9가지 범주로 확장하여 기획 단계에서부터 이를 고려한 개발을 통해 게임, 전시, 공연, 관광, 디자인, 에듀테크, 미디어 등에서 감성 콘텐츠 개발에 활용될 수 있음을 제안한다. 후속 연구를 통하여 더욱 다양한 일상 대화들을 추가로 수집하여 감정분석을 진행한다면 더욱 정확한 결과를 얻을 수 있을 것으로 기대된다.

Keywords

References

  1. D, H. Jo, I. Jeon and M. K. Moon. "Scenario Generation Assistance System Using GPT-3," Proceedings of the Korean Society of Computer Information Conference(KSCI), Vol.30, No.2, pp. 503-504. July 2022.
  2. W. S. Sim, Y. H. Moon and H. B. Seo, "Improvement of Online-Shopping Mall Review System using Natural Language Processing," Electrical Machinery & Energy Conversion Systems Society Annual Spring Conference (KIEE), pp.217-218, 2022.
  3. J. Garten, K. Sagae and V. Ustun, "Combining distributed vector representations for words," In Proceedings of the 1st workshop on vector space modeling for natural language processing (NAACL-HLT), pp. 95-101, June 2015. DOI: 10.3115/v1/W15-1513
  4. A. Karpathy, J. Johnson and L. Fei-Fei, "Visualizing and Understanding Recurrent Networks," arXiv:1506.02078v2, 2015. DOI: 10.48550/arXiv.1506.02078
  5. K. H. Cho, B. van Merrienboer and C. Gulcehre, "Learning phrase representations using RNN encoder-decoder for statistical machine translation," Conference on Empirical Methods in Natural Language Processing(EMNLP), pp.1724-1734, 2014. DOI: 10.3115/v1/D14-1179
  6. P. Ekman, "An argument for basic emotions. Cognition and Emotion 6," Congnition and Emotion, Vol. 3, No. 4, pp. 169-200, Oct 1992. DOI: /10.1080/02699939208411068
  7. A. Vaswani, N. Shazeer and N. Parmar, "Attention is all you need," Advances in Neural Information Processing Systems 30, pp. 6000-6010, Dec 2017.
  8. Z. Hwang, "Understanding and Trends of a Transformer Deep Learning Model in the Liberal Arts Education," The Journal of Liberal Arts (HILAS), Vol.9, No.2, pp. 145-156, 2022. DOI: 10.55450/hilas.2022.9.2.145
  9. T. Brown, "Language models are few-shot learners," Advances in Neural Information Processing Systems 3 (NeurlPS), pp. 1877-1901, 2020. DOI: 10.48550/arXiv.2005.14165
  10. H. G. Jeon, K. C. Jung and K. A. Kwon, "Machine Learning Language Model Implementation Using Literary Texts," The Journal of the Convergence on Culture Technology(JCCT), Vol.7 No.2, pp. 427-436, 2020. DOI : 10.17703/JCCT.2020.6.2.449
  11. Jae-Kyung Sung, Yung-Bok Kim and Yong-Guk Kim, "Deep learning-based Multilingual Sentimental Analysis using English Review Data," The Journal of the Institute of Internet (JIIBC), Broadcasting and Communication, Vol.19, No.3, pp.9-15, 2019. DOI : 10.7236/JIIBC.2019.19.3.9
  12. Hyorim Shin, Junho Choi, "Analysis of User Reviews for Webtoon Applications Using Text Mining," The Journal of the Convergence on Culture Technology(JCCT), Vol.8 No.4, pp. 457-468, 2022. DOI : 10.17703/JCCT.2022.8.4.457
  13. S. S. Na, J. Y. Kang and S. J. Rlee, "Implementation of Age Prediction Model for Elementary and Secondary Learners' Writings Using Transformer Language Model," Journal of CheongRam Korean Language Education, No.90, pp. 51-96, 2022. DOI : 10.26589/jockle..90.202211.51
  14. S. I. Kim, D. S. Kim and J. W. Kim, "Public Sentiment Analysis of Korean Top-10 Companies : Big Data Approach Using Multi-categorical Sentiment Lexicon," Journal of Intelligence and Information System(JIIS), Vol.22, No.3, pp.45-69, 2016. DOI : 10.13088/jiis.2016.22.3.045