Sentiment Analysis of Skincare Live Broadcasts: A RoBERTa-Based Text Mining Approach
Keywords:
Sentiment Analysis, RoBERTa, Live Commerce, Text Mining, SkincareAbstract
The rapid rise of social commerce, particularly TikTok Live, has transformed the Indonesian
skincare industry. During these live broadcasts, viewers generate massive amounts of
unstructured text in the comment section. While these comments contain valuable insights into
consumer behavior, they are difficult to analyze manually due to non-standard slang and high
velocity. This study aims to classify audience sentiment polarity and extract dominant
conversational themes from a skincare brand's TikTok Live broadcast. The methodology involved
scraping 693 real-time comments, conducting rigorous text preprocessing, and applying a fine-
tuned RoBERTa transformer model. To overcome the common misclassification of consumer
inquiries, a rule-based algorithm was integrated to assign interrogative comments as 'neutral
automatically'. Visualizations were generated using pie charts and N-Gram Word Clouds. The
findings challenge the assumption that comment sections are primarily evaluative spaces. Results
showed an overwhelming dominance of neutral sentiment (75.7%), indicating that consumers use
the platform primarily as a virtual consultation clinic. Negative sentiment (21.5%) revealed
critical consumer pain points, driven by dermatological insecurities (e.g., acne) and transactional
frictions like the unavailability of Cash on Delivery (COD). Positive sentiment (2.8%) was largely
driven by perceived product safety and promotional bundles. Managerially, these insights suggest
that brands must train live hosts as knowledgeable beauty advisors to effectively address real-
time consultations. Future research should expand rule-based slang dictionaries and conduct
domain-specific fine-tuning to handle linguistic anomalies in Shoppertainment platforms better.
References
[1] R. Schmitt, “From raw text to fairseq RoBERTa: A modular snakemake-based framework
enabling language-specific BPE tokenization,” Software Impacts, vol. 27, p. 100824, Apr.
2026, doi: 10.1016/J.SIMPA.2026.100824.
[2] G. Gelastopoulos and C. Keramydas, “A systematic review of text mining analytics for
supply chain risk management using online data,” Supply Chain Analytics, vol. 12, p.
100167, Dec. 2025, doi: 10.1016/J.SCA.2025.100167.
[3] Y. Li and C. W. Zobel, “Exploring supply chain network resilience in the presence of the
ripple effect,” Int. J. Prod. Econ., vol. 228, Oct. 2020, doi: 10.1016/j.ijpe.2020.107693.
[4] M. Pontiki et al., “SemEval-2016 task 5: Aspect based sentiment analysis,” SemEval 2016
- 10th International Workshop on Semantic Evaluation, Proceedings, pp. 19–30, 2016,
doi: 10.18653/V1/S16-1002.
[5] D. Tang, F. Wei, N. Yang, M. Zhou, T. Liu, and B. Qin, “Learning sentiment-specific
word embedding for twitter sentiment classification,” 52nd Annual Meeting of the
Association for Computational Linguistics, ACL 2014 - Proceedings of the Conference,
vol. 1, pp. 1555–1565, 2014, doi: 10.3115/V1/P14-1146.
[6] E. Cambria, “Affective Computing and Sentiment Analysis,” IEEE Intell. Syst., vol. 31,
no. 2, pp. 102–107, Mar. 2016, doi: 10.1109/MIS.2016.31.
[7] Y. Song, J. Wang, T. Jiang, Z. Liu, and Y. Rao, “Attentional Encoder Network for
Targeted Sentiment Classification,” Apr. 2019, doi: 10.1007/978-3-030-30490-4_9.
[8] L. You, F. Han, J. Peng, H. Jin, and C. Claramunt, “ASK-RoBERTa: A pretraining model
for aspect-based sentiment classification via sentiment knowledge mining,” Knowl. Based.
Syst., vol. 253, p. 109511, Oct. 2022, doi: 10.1016/J.KNOSYS.2022.109511.
[9] Y. Liu et al., “RoBERTa: A Robustly Optimized BERT Pretraining Approach,” Jul. 2019,
Accessed: Apr. 21, 2026. [Online]. Available: http://arxiv.org/abs/1907.11692
[10] P. Nandwani and R. Verma, “A review on sentiment analysis and emotion detection from
text,” Soc. Netw. Anal. Min., vol. 11, no. 1, Dec. 2021, doi: 10.1007/S13278-021-00776-
6.
[11] X. Fang and J. Zhan, “Sentiment analysis using product review data,” J. Big Data, vol. 2,
no. 1, Dec. 2015, doi: 10.1186/S40537-015-0015-2.[12] M. Rodríguez-Ibánez, A. Casánez-Ventura, F. Castejón-Mateos, and P. M. Cuenca-
Jiménez, “A review on sentiment analysis from social media platforms,” Expert Syst.
Appl., vol. 223, Aug. 2023, doi: 10.1016/j.eswa.2023.119862.
[13] S. Sindhu, S. Kumar, and A. Noliya, “A Review on Sentiment Analysis using Machine
Learning,” International Conference on Innovative Data Communication Technologies
and Application, ICIDCA 2023 - Proceedings, pp. 138–142, 2023, doi:
10.1109/ICIDCA56705.2023.10099665.
[14] B. Paneru, B. Thapa, and B. Paneru, “Sentiment analysis of movie reviews: A flask
application using CNN with RoBERTa embeddings,” Systems and Soft Computing, vol.
7, p. 200192, Dec. 2025, doi: 10.1016/J.SASC.2025.200192.
[15] M. G. Yazdi, H. Rafieizadeh, and P. Tajasob, “Sentiment analysis of the 2024 Paris
Olympics using RoBERTa language model,” International Journal of Event and Festival
Management, vol. 17, no. 1, pp. 133–160, Sep. 2025, doi: 10.1108/IJEFM-01-2025-0007.
[16] J. Dai, Y. Zhao, and Z. Li, “Sentiment-topic dynamic collaborative analysis-based public
opinion mapping in aviation disaster management: A case study of the MU5735 air crash,”
International Journal of Disaster Risk Reduction, vol. 102, Feb. 2024, doi:
10.1016/j.ijdrr.2024.104268.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Muhammad Imam Baihaqi, Ahmad Arro’uf Sulfuadi (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
