A systemized knowledge-enhanced approach for sentiment analysis of social media cosplay video comments to optimize cultural dissemination strategies
The niche terminology and cultural symbols used in social media cosplay video comments are difficult to accurately identify with general sentiment analysis tools, leading to imprecise audience sentiment feedback. Therefore, this paper proposes a systematic, knowledge-enhanced approach, cosplay knowledge-enhanced bidirectional encoder representation from transformers (Cos-BERT), based on domain-specific terminology injection. By constructing a domain terminology system that integrates cultural symbols and sentiment tags from specific social circles, positive and negative dictionaries of social circle terms and cultural symbols were extracted. For each comment text, the match count across the three dictionaries was calculated and normalized into a knowledge feature vector. Simultaneously, a BERT encoder was used to extract the semantic representation of the comment text. The knowledge feature vector and semantic representation were concatenated and input into the classification layer to achieve fine-grained discrimination of positive, negative, and neutral audience sentiment. Experimental results show that the proposed Cos-BERT model achieves 0.86 accuracy on the test set, with a macro-average F1 score of 0.85. Furthermore, all three dictionary modules—positive, negative, and cultural symbol—contribute positively to sentiment discrimination. This systematic approach provides a technical pathway for extracting the audience’s genuine emotional responses to cultural symbols from massive amounts of comments.
Al-Tameemi, I. K. S., Feizi-Derakhshi, M. R., Pashazadeh, S., & Asadpour, M. (2024). A comprehensive review of visual–textual sentiment analysis from social media networks. Journal of Computational Social Science, 7(3), 2767-2838. https://doi.org/10.1007/s42001-024-00326-y
Bao, L., Lambert, P., & Badia, T. (2024). Improving aspect-based neural sentiment classification with lexicon enhancement, attention regularization and sentiment induction. Natural Language Engineering, 30(1), 1-30. https://doi.org/10.1017/S1351324922000432
Bhatia, K. V., Pathak-Shelat, M., Sinha, S., & Mishra, T. (2025). Global influencers’ content creation strategies: Negotiating with platform affordances to practice vernacular creativity. Media, Culture & Society, 47(1), 130-153. https://doi.org/10.1177/01634437241276408
Bonifazi, G., Cauteruccio, F., Corradini, E., Marchetti, M., Terracina, G., Ursino, D., & Virgili, L. (2023). A framework for investigating the dynamics of user and community sentiments in a social platform. Data & Knowledge Engineering, 146(1), 102183-102196. https://doi.org/10.1016/j.datak.2023.102183
Bucur, A. M., Chulvi, B., Cosma, A., & Rosso, P. (2024). The expression of happiness in social media of individuals reporting depression. IEEE Transactions on Affective Computing, 16(1), 360-375. https://doi.org/10.1109/TAFFC.2024.3434482
Cao, D., Tang, X., Lin, Y., & Lin, D. (2025). Short video rumor detection based on causal graph. Information Sciences, 703(3), 121941-121956. https://doi.org/10.1016/j.ins.2025.121941
Dong, Z. (2025). Exploring cross-cultural communication content adaptability through advanced natural language processing and sentiment analysis. Systems and Soft Computing, 7(2), 200290-200296. https://doi.org/10.1016/j.sasc.2025.200290
Dui, Y., & Hu, H. (2024). Social media public opinion detection using multimodal natural language processing and attention mechanisms. IET Information Security, 2024(1), 1-13. https://doi.org/10.1049/2024/8880804
Durga, P., & Godavarthi, D. (2023). Deep-sentiment: An effective deep sentiment analysis using a decision-based recurrent neural network (D-RNN). IEEE Access, 11(2), 108433-108447. https://doi.org/10.1109/ACCESS.2023.3320738
Feng, W., Wang, X., Cao, D., & Lin, D. (2024). An autoencoder-based self-supervised learning for multimodal sentiment analysis. Information Sciences, 675(12), 120682-120695. https://doi.org/10.1016/j.ins.2024.120682
Guo, Z., Jia, Q., Fan, B., Wang, D., Xu, C., Wang, Y., Zhao, Y., & Li, R. (2024). MVIndEmo: a dataset for micro video public-induced emotion prediction on social media. Multimedia Systems, 30(1), 58-66. https://doi.org/10.1007/s00530-023-01221-8
Guo, Q., & Guo, Y. (2022). Lexicon enhanced Chinese named entity recognition with pointer network. Neural Computing and Applications, 34(17), 14535-14555. https://doi.org/10.1007/s00521-022-07287-1
Hou, X. D., Guntuku, S. C., Cho, Y. M., Sherman, G., Zhang, T., Li, M., Ungar, L., Tay, L., & Chen, T. (2024). A cross-cultural examination of temporal orientation through everyday language on social media. Plos one, 19(3), e0292963-e0292976. https://doi.org/10.1371/journal.pone.0292963
Hwang, Y., Kim, C., Park, G., & Kwon, H. Y. (2025). Affective audio feature-based multimodal learning model for extracting video highlights. Applied Soft Computing, 12(3), 113600-113612. https://doi.org/10.1016/j.asoc.2025.113600
Ignacio, M. J., Nguyen, T. T., Jin, H., & Kim, Y. G. (2024). Meme analysis using LLM-based contextual information and u-net encapsulated transformer. IEEE Access, 12(3), 125993-126005. https://doi.org/10.1109/ACCESS.2024.3424883
Kridera, S., & Kanavos, A. (2025). Modelling Social Attachment and Mental States from Facebook Activity with Machine Learning. Information, 16(9), 772-789. https://doi.org/10.3390/info16090772
Levy, S., Gvili, Y., & Hino, H. (2025). The effects of social capital on cross-cultural eWOM communication in minority-mainstream consumer interactions on social media. International Journal of Emerging Markets, 20(7), 3122-3144. https://doi.org/10.1108/IJOEM-03-2023-0389
Long, J., Li, Z., Xuan, Q., Fu, C., Peng, S., & Min, Y. (2023). Social media opinion analysis model based on fusion of text and structural features. Applied Sciences, 13(12), 7221-7236. https://doi.org/10.3390/app13127221
Ma, J. (2026). Multimodal emotion feature extraction and information fusion methods for video content. International Journal of Information and Communication Technology, 27(28), 43-59. https://doi.org/10.1504/IJICT.2026.152548
Ma, Y., & Li, M. (2024). Unpacking translanguaging in refusals on Chinese social media: Strategies, distribution, and functions. Multilingua, 43(4), 487-524. https://doi.org/10.1515/multi-2023-0071
Moura, B. M., & de Souza-Leão, A. L. M. (2025). Courage to be a fan: the alethurgical dispositif that governs cosplay. Revista FAMECOS, 32(1), e46784-e46784. https://doi.org/10.15448/1980-3729.2025.1.46784
Nafis, A., & Ghazali, M. (2025). Cross-cultural emotion analysis on X using bidirectional encoder representations from transformers (BERT). Journal of Engineering and Technology (JET), 16(1), 77-94. https://doi.org/10.54554/jet.2025.16.1.007
Noman, M., Gurgun, S., Phalp, K., & Ali, R. (2024). Designing social media to foster user engagement in challenging misinformation: a cross-cultural comparison between the UK and Arab countries. Humanities and Social Sciences Communications, 11(1), 12-26. https://doi.org/10.1057/s41599-024-03524-1
Ocal, A. (2026). A Computational Analysis of Emotions and Topics in YouTube Discourse on Sora. Applied Sciences, 16(5), 2519-2529. https://doi.org/10.3390/app16052519
Putri, S. A., & Amiralda, R. P. (2024). How Does a Cosplayer Integrate Personal Identity with the Personality of Their Chosen Character?(Cosplay Events after the Pandemic in Jakarta). Wacana Journal of Social and Humanity Studies, 27(3), 132-140. https://doi.org/10.21776/ub.wacana.2024.027.03.06
Shi, W., Zhang, J., & He, S. (2024). Understanding public opinions on Chinese short video platform by multimodal sentiment analysis using deep learning-based techniques. Kybernetes, 53(12), 5932-5950. https://doi.org/10.1108/K-04-2023-0723
Subbaiah, B., Murugesan, K., Saravanan, P., & Marudhamuthu, K. (2024). An efficient multimodal sentiment analysis in social media using hybrid optimal multi-scale residual attention network. Artificial Intelligence Review, 57(2), 34-45. https://doi.org/10.1007/s10462-023-10645-7
Sun, H., Liu, J., Chen, Y. W., & Lin, L. (2023). Modality-invariant temporal representation learning for multimodal sentiment classification. Information Fusion, 91(2), 504-514. https://doi.org/10.1016/j.inffus.2022.10.031
Vardell, E., Thomas, P., & Wang, T. (2023). The informational “cosplay journey” of Star Wars cosplayers in the context of a Facebook group. Information Research an international electronic journal, 28(3), 37-58. https://doi.org/10.47989/ir283199
Watimin, N. H., Zanuddin, H., Rahamad, M. S., & Yadegaridehkordi, E. (2023). Content framing role on public sentiment formation for pre-crisis detection on sensitive issue via sentiment analysis and content analysis. Plos one, 18(10), 1-20. https://doi.org/10.1371/journal.pone.0287367
Wei, Q., Zhou, Y., Xiang, S., Xiao, L., & Zhang, Y. (2024). MEAS: Multimodal Emotion Analysis System for Short Videos on Social Media Platforms. IEEE Transactions on Computational Social Systems, 12(5), 2398-2410. https://doi.org/10.1109/TCSS.2024.3490846
Yakut, S., Tuten, Y. T., Caglar, E., & Aktas, M. S. (2026). A Multimodal Transformer-Based Framework for Emotion Analysis in Multilingual Video Content. Computers, 15(2), 77-88. https://doi.org/10.3390/computers15020077
Yang, K., & Hau, T. S. (2025). “I Was Happy the Whole Time and Still Feel Happy About It Now”: Online and Offline Experiences of Hong Kong Cosplayers. Journal of Asia-Pacific Pop Culture, 10(2), 166-191. https://doi.org/10.5325/jasiapacipopcult.10.2.0166
Zhang, Q., Kadir, Z. A., & Wan, N. W. M. (2025). Cross-Cultural Emotional Polarity Comparison of Emoji on Chinese and English Social Media in the Context of the World Cup. ESTEEM Journal of Social Sciences and Humanities, 9(2), 14-22. https://doi.org/10.24191/ejssh.v9i2.6850
Zhang, X. (2024). Intercultural attitudes embedded in microblogging: sentiment and content analyses of data from Sina Weibo. Journalism and Media, 5(4), 1477-1493. https://doi.org/10.3390/journalmedia5040092