ESP-4081

Published

2025-09-24

Issue

Vol. 10 No. 9 (2025): Published

Section

Research Articles

License

Copyright (c) 2025 Zhixiang Xu, Adzrool Idzwan bin Ismail

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

The journal adopts the Attribution-NonCommercial 4.0 International (CC BY-NC 4.0), which means that anyone can reuse and redistribute the materials for non-commercial purposes as long as you follow the license terms and the original source is properly cited.

Author(s) shall retain the copyright of their work and grant the Journal/Publisher rights for the first publication with the work concurrently licensed since 2023 Vol.8 No.2.

Under this license, author(s) will allow third parties to download, reuse, reprint, modify, distribute and/or copy the content under the condition that the authors are given credit. No permission is required from the authors or the publisher.

This broad license intends to facilitate free access, as well as the unrestricted use of original works of all types. This ensures that the published work is freely and openly available in perpetuity.

By providing open access, the following benefits are brought about:

  • Higher Visibility, Availability and Citations-free and unlimited accessibility of the publication over the internet without any restrictions increases citation of the article.
  • Ease of search-publications are easily searchable in search engines and indexing databases.
  • Rapid Publication – accepted papers are immediately published online.
  • Available for free download immediately after publication at https://esp.as-pub.com/index.php/ESP

 

Copyright Statement

1.The authors certify that the submitted manuscripts are original works, do not infringe the rights of others, are free from academic misconduct and confidentiality issues, and that there are no disputes over the authorship scheme of the collaborative articles. In case of infringement, academic misconduct and confidentiality issues, as well as disputes over the authorship scheme, all responsibilities will be borne by the authors.

2. The author agrees to grant the Editorial Office of Environment and Social Psychology a licence to use the reproduction right, distribution right, information network dissemination right, performance right, translation right, and compilation right of the submitted manuscript, including the work as a whole, as well as the diagrams, tables, abstracts, and any other parts that can be extracted from the work and used in accordance with the characteristics of the journal. The Editorial Board of Environment and Social Psychology has the right to use and sub-licence the above mentioned works for wide dissemination in print, electronic and online versions, and, in accordance with the characteristics of the periodical, for the period of legal protection of the property right of the copyright in the work, and for the territorial scope of the work throughout the world.

3. The authors are entitled to the copyright of their works under the relevant laws of Singapore, provided that they do not exercise their rights in a manner prejudicial to the interests of the Journal.

About Licence

Environment and Social Psychology is an open access journal and all published work is available under the Creative Commons Licence, Authors shall retain copyright of their work and grant the journal/publisher the right of first publication, and their work shall be licensed under the Attribution-NonCommercial 4.0 International (CC BY-NC 4.0).

Under this licence, the author grants permission to third parties to download, reuse, reprint, modify, distribute and/or copy the content with attribution to the author. No permission from the author or publisher is required.

This broad licence is intended to facilitate free access to and unrestricted use of original works of all kinds. This ensures that published works remain free and accessible in perpetuity. Submitted manuscripts, once accepted, are immediately available to the public and permanently accessible free of charge on the journal’s official website (https://esp.as-pub.com/index.php/ESP). Allowing users to read, download, copy, print, search for or link to the full text of the article, or use it for other legal purposes. However, the use of the work must retain the author's signature, be limited to non-commercial purposes, and not be interpretative.

Click to download <Agreement on the Licence for the Use of Copyright on Environmental and Social Psychology>.

How to Cite

Zhixiang Xu, & Adzrool Idzwan bin Ismail. (2025). Generative AI-based dialogue generation and optimization in the style of traditional Chinese opera Dream of the Red Chamber using K-means Clustering. Environment and Social Psychology, 10(9), ESP-4081. https://doi.org/10.59429/esp.v10i9.4081

Generative AI-based dialogue generation and optimization in the style of traditional Chinese opera Dream of the Red Chamber using K-means Clustering

Zhixiang Xu

chool of Creative Industry Management & Performing Art (SCIMPA), Universiti Utara Malaysia, 06010 Kedah Darul Aman, Malaysia ; Xi’an International University, Xi’an, 710077, Shaanxi, China

Adzrool Idzwan bin Ismail

chool of Creative Industry Management & Performing Art (SCIMPA), Universiti Utara Malaysia, 06010 Kedah Darul Aman, Malaysia


DOI: https://doi.org/10.59429/esp.v10i9.4081


Keywords: Generative AI; Classical Chinese opera; GPT-4o; text generation; Dream of the Red Chamber

Abstract

This study proposes a generative artificial intelligence framework that integrates prompt engineering and K-means clustering to generate stylized dialogues from the tradition of Dream of the Red Chamber. The prompts for each character were constructed from the linguistic features of classical texts, and each generated sentence was encoded into semantic, lexical, and keyword-based features. K-means clustering was used to identify stylistic categories, and the number of clusters was validated using the Elbow Method and the Silhouette Coefficient. Evaluation used BLEU, Perplexity, and a custom Style Consistency Score. As a result, the K-means-enhanced model improved fluency and style consistency over the baseline. This method provides a quantitative approach to improving generative models in traditional cultural domains.

References

[1]. 1.Wang, Z., Xu, Y., Zhang, L., & Chen, H. (2022). Classical Chinese poetry generation with pretrained language models. arXiv:2211.02541. https://arxiv.org/abs/2211.02541

[2]. 2.K. Davis and T. Nguyen, Style Transfer in AI-Generated Dialogues, Computational Linguistics, vol. 48, no. 2, pp. 289–305, 2022.

[3]. 3.P. Gonzalez and H. Kim, Perplexity metrics in AI dialogue systems, Journal of Computational Science, vol. 39, p. 101123, 2022.

[4]. 4.X. Li, J. Wang, and Y. Zhang, Aesthetic Rhythm in Traditional Chinese Opera Adaptation, Journal of Chinese Literature and Art, vol. 32, no. 2, pp. 85–101, 2020.

[5]. 5.X. Li, H. Sun, and Z. Zhao, Stylistic Evaluation of Chinese Classical Texts with Neural Models, Natural Language Engineering, vol. 28, no. 5, pp. 593–610, 2022.

[6]. 6.X. Liu, P. Li, M. Ding, Z. Liu, and J. Tang, Prompt learning for text generation: A survey, ACM Transactions on Intelligent Systems and Technology, vol. 13, no. 4, Article 41, 2022.

[7]. 7.L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. Wainwright, P. Mishkin, and P. Christiano, Training language models to follow instructions with human feedback, arXiv preprint, 2022.

[8]. 8.K. Papineni, S. Roukos, T. Ward, and W. J. Zhu, BLEU: A method for automatic evaluation of machine translation, in Proc. 40th Annual Meeting of the Association for Computational Linguistics, pp. 311–318, 2002.

[9]. 9.J. Wang and Y. Zhang, Unsupervised Stylistic Clustering for Chinese Text Generation, in Proc. 59th Annual Meeting of the Association for Computational Linguistics, pp. 1432–1445, 2021.

[10]. 10.Y. Xie, J. Li, and Z. Zhou, Classical Chinese Text Generation Using Pre-trained Language Models, Journal of Computational Linguistics, vol. 47, no. 3, pp. 425–448, 2021.

[11]. 11.X. Zhang and P. Sun, Stylistic control in Chinese classical text generation with prompt-based models, in Proc. 61st Annual Meeting of the Association for Computational Linguistics, pp. 1243–1255, 2023.

[12]. 12.Y. Bai and S. Lei, Cross-language dissemination of Chinese classical literature using multimodal deep learning and artificial intelligence, Scientific Reports, vol. 15, no. 1, p. 13855, Jul. 2025.

[13]. 13.Y. Zhang, J. Wu, and K. Li, Prompt framework for role-playing: Generation and persona consistency in large language models, arXiv preprint arXiv:2406.00627, Jun. 2024.

[14]. 14.T. Joseph and H. K. Male, Exploring the synergy of grammar-aware prompt engineering and formal methods for mitigating hallucinations in LLMs, East African Journal of Information Technology, vol. 7, no. 1, pp. 188–201, Aug. 2024.

[15]. 15.Z. Chen, A polishing model for machine-generated ancient Chinese poetry, Neural Processing Letters, vol. 56, article 77, Mar. 2024.

[16]. 16.K. Ji, Enhancing persona consistency for LLMs’ role-playing, in Findings of the Association for Computational Linguistics (ACL), 2025.

[17]. 17.D. Liu, A Persona-Aware LLM-Enhanced Framework for Multi-Session Personalized Dialogue Generation (PALACE), in Findings of the Association for Computational Linguistics (ACL), 2025.

[18]. 18.B. Li, A systematic review of empirical generative AI research in language learning and teaching, Computers & Education: Artificial Intelligence, vol. 6, no. 100215, 2025.

[19]. 19.R. He, Generative artificial intelligence: A historical perspective, National Science Review, vol. 12, no. 5, nwaf050, 2025.

[20]. 20.G. Schryen, Exploring the scope of generative AI in literature review, Electronic Markets, 2025.

[21]. 21.A. Adel, Can generative AI reliably synthesise literature? Exploring the accuracy and risks of ChatGPT, AI & Society, 2025.

[22]. 22.J. Zheng, M. Wang, and K. Ren, When ‘A Helpful Assistant’ Is Not Really Helpful: Personas in System Prompts Do Not Improve Performance in LLMs, arXiv preprint arXiv:2311.10054, 2023.

[23]. 23.J. Kim, N. Yang, and K. Jung, Persona is a double-edged sword: Mitigating the negative impact of role-playing prompts in LLMs, arXiv preprint arXiv:2408.08631, 2024.

[24]. 24.F. A. Tan, L. Zhao, and P. Huang, PHAnToM: Persona-based prompting has an effect on Theory-of-Mind reasoning in LLMs, arXiv preprint arXiv:2403.02246, 2024.

[25]. 25.The Authors Guild, Legal tussle between authors and AI: Class action lawsuit against OpenAI for unauthorized use of books, Vanity Fair, Oct. 2023.



21 Woodlands Close #02-10, Primz Bizhub,Postal 737854, Singapore

Email:editorial_office@as-pub.com