İran–İsrail 12 Gün Savaşı Sırasında Çok Hedefli Duruş Sınıflandırması Yoluyla Kamusal Söylemin Olay Penceresi Analizi

Abstract

Bu çalışma, İran–İsrail 12 gün savaşı sürecinde çevrim içi izleyicilerin ABD’nin saldırıları, İsrail ve İran olmak üzere birbiriyle ilişkili üç hedefe yönelik duruşlarını ve yorumlayıcı çerçevelerini, önde gelen ABD haber videolarındaki YouTube yorumları üzerinden analiz etmektedir. 112.000 yorumdan oluşan derlem, olay öncesi, olay penceresi ve olay sonrası dönemlere ayrılmıştır. Duruş, her hedef için ayrı ayrı Destek, Karşıtlık, Belirsiz ve Uygulanamaz olmak üzere dört etiket kullanılarak modellenmiş ve böylece değerlendirme yönü ile hedefin söylemdeki görünürlüğünün birbirinden ayrıştırılması mümkün olmuştur. İnsanlar tarafından etiketlenen ve çift kodlanan 2.500 yorumluk örneklem, anotasyon güvenilirliği testleri ve model eğitimi için kullanılmıştır. Sınıflandırıcılar, video örtüşmesini kontrol eden ve zamansal genelleştirilebilirliği dikkate alan veri sızıntısına dirençli protokollerle değerlendirilmiş, ardından tüm derleme uygulanmıştır. Bu değerlendirme koşullarında, TF-IDF ile Lojistik Regresyon modeli ortalama dört sınıflı Macro-F1 performansında en yüksek sonucu elde etmiş ve özellikle zamansal genelleme açısından hiyerarşik DeBERTa modelinden daha güçlü performans göstermiştir. Bulgular, tartışmanın çoğunlukla ABD saldırılarına odaklandığını; İsrail ve İran’ın ise daha seyrek biçimde doğrudan değerlendirildiğini göstermektedir. Olay penceresi sırasında, uygulanabilir yorumlar arasında ABD saldırılarına yönelik destek artmıştır. Konu modellemesi, tırmanma, meşruiyet, hesap verebilirlik ve bölgesel güvenlik gibi tekrar eden çerçeveleri ortaya koymuştur

Keywords

• Duruş Tespiti
• Olay Merkezli Analiz
• Youtube Yorumları
• İran–İsrail Savaşı
• ABD Saldırıları
• Çerçeveleme
• BERTopic

Event-Window Analysis of Public Discourse via Multi-Target Stance Classification During the 12-Day Iran–Israel War

Abstract

This study analyzes how online audiences expressed stance and interpretive frames toward three related targets, U.S. strikes, Israel, and Iran, during the 12-day Iran–Israel war period through YouTube comments on prominent U.S. news videos. A corpus of 112,000 comments was divided into pre-event, event-window, and post-event periods. Stance was modeled separately for each target using four labels: Support, Oppose, Unclear, and Not Applicable, allowing evaluation and target salience to be distinguished. A humanlabeled, double-coded sample of 2,500 comments supported annotation reliability testing and model training. Classifiers were evaluated with leakage-resistant protocols controlling for video overlap and temporal generalization, then applied to the full corpus. Under these evaluation settings, TF-IDF with Logistic Regression achieved the strongest average four-way Macro-F1, outperforming the hierarchical DeBERTa model, particularly under temporal generalization. Results show that discussion most often focused on U.S. strikes, while Israel and Iran were less frequently evaluated directly. During the event window, support for U.S. strikes increased among applicable comments. Topic modeling identified recurring frames of escalation, legitimacy, accountability, and regional security.

Keywords

• Stance Detection
• Event Centered Analysis
• Youtube Comments
• Iran–Israel War
• U.S. Strikes
• Framing
• BERTopic

References

1. ALDayel, A., & Magdy, W. (2021). Stance detection on social media: State of the art and trends. Information Processing & Management, 58(4), 102597. https://doi.org/10.1016/j.ipm.2021.102597
2. Aziz, R. A., Agarwal, P., McGuinness, J., Karmaker, C. L., & Zhuang, J. (2024). Temporal patterns and life cycle dynamics of social media user activity duringdisasters: A data-driven approach for effective crisis communication. Expert Systems with Applications, 255, 124454. https://doi.org/10.1016/j.eswa.2024.124454
3. Berg, J., Fagerholm, A., & Strandberg, K. (2026). Quality User-Generated Content? A Case Study of the Quality of Online News Comments on the Site of Finnish Public Service Broadcaster Yle. Journalism Practice, 20(1), 338–359. https://doi.org/10.1080/17512786.2024.2329759
4. Berk, N. (2025). The Impact of Media Framing in Complex Information Environments. Political Communication, 42(5), 757–773. https://doi.org/10.1080/10584609.2025.2456519
5. Cakmak, M. C., Agarwal, N., & Oni, R. (2024). The bias beneath: Analyzing drift in YouTube’s algorithmic recommendations. Social Network Analysis and Mining, 14(1), 171. https://doi.org/10.1007/s13278-024-01343-5
6. Chagnon, E., Pandolfi, R., Donatelli, J., & Ushizima, D. (2024). Benchmarking topic models on scientific articles using BERTeley. Natural Language Processing Journal, 6, 100044. https://doi.org/10.1016/j.nlp.2023.100044
7. Cheddak, A., Ait Baha, T., Es-Saady, Y., El Hajji, M., & Baslam, M. (2024). BERTopic for Enhanced Idea Management and Topic Generation in Brainstorming Sessions. Information, 15(6), 365. https://doi.org/10.3390/info15060365
8. Chen, K., Zhang, Z., Long, J., & Zhang, H. (2016). Turning from TF-IDF to TF-IGM for term weighting in text classification. Expert Systems with Applications, 66, 245–260. https://doi.org/10.1016/j.eswa.2016.09.009

9. Chen, Y., Kmetty, Z., Iñiguez, G., & Omodei, E. (2025). The public that engages invisibly: What visible engagement fails to capture in online political communication. Communication Methods and Measures, 19(4), 294–312. https://doi.org/10.1080/19312458.2025.2542725
10. Dubovi, I., & Tabak, I. (2020). An empirical analysis of knowledge co-construction in YouTube comments. Computers & Education, 156, 103939. https://doi.org/10.1016/j.compedu.2020.103939
11. Duran, N., Battle, S., & Smith, J. (2022). Inter-annotator Agreement Using the Conversation Analysis Modelling Schema, for Dialogue. Communication Methods and Measures, 16(3), 182–214. https://doi.org/10.1080/19312458.2021.2020229
12. Eisele, O., Heidenreich, T., Litvyak, O., & Boomgaarden, H. G. (2023). Capturing a News Frame – Comparing Machine-Learning Approaches to Frame Analysis with Different Degrees of Supervision. Communication Methods and Measures, 17(3), 205–226. https://doi.org/10.1080/19312458.2023.2230560
13. Eldridge II, S. A., & Franklin, B. (Eds.). (2016). The Routledge Companion to Digital Journalism Studies (1st ed.). Routledge. https://doi.org/10.4324/9781315713793
14. Fenza, G., Gallo, M., Loia, V., Petrone, A., & Stanzione, C. (2023). Concept-drift detection index based on fuzzy formal concept analysis for fake news classifiers. Technological Forecasting and Social Change, 194, 122640. https://doi.org/10.1016/j.techfore.2023.122640
15. Flores, P. M., & Hilbert, M. (2023). Temporal communication dynamics in the aftermath of large-scale upheavals: Do digital footprints reveal a stage model? Journal of Computational Social Science, 6(2), 973–999. https://doi.org/10.1007/s42001-023-00218-7
16. Garg, K., & Caragea, C. (2024). Stanceformer: Target-Aware Transformer for Stance Detection. In Y. Al-Onaizan, M. Bansal, & Y.-N. Chen (Eds.), Findings of the Association for Computational Linguistics: EMNLP 2024 (pp. 4969–4984). Association for Computational Linguistics. https://doi.org/10.18653/v1/2024.findings-emnlp.286
17. Gómez-Suta, M., Echeverry-Correa, J., & Soto-Mejía, J. A. (2023). Stance detection in tweets: A topic modeling approach supporting explainability. Expert Systems with Applications, 214, 119046. https://doi.org/10.1016/j.eswa.2022.119046
18. Grootendorst, M. (2022). BERTopic: Neural topic modeling with a class-based TF-IDF procedure. arXiv. https://arxiv.org/abs/2203.05794
19. He, P., Gao, J., & Chen, W. (2023). DeBERTaV3: Improving DeBERTa using ELECTRAStyle Pre-Training with Gradient-Disentangled Embedding Sharing. arXiv. https://arxiv.org/abs/2111.09543
20. Heuer, H., Hoch, H., Breiter, A., & Theocharis, Y. (2021). Auditing the Biases Enacted by YouTube for Political Topics in Germany. Proceedings of Mensch Und Computer 2021, MuC’21, 456–468. https://doi.org/10.1145/3473856.3473864
21. Jamadi Khiabani, P., & Zubiaga, A. (2025). Cross-target stance detection: A survey of techniques, datasets, and challenges. Expert Systems with Applications, 283, 127790. https://doi.org/10.1016/j.eswa.2025.127790
22. Khan, M. L. (2017). Social media engagement: What motivates user participation and consumption on YouTube? Computers in Human Behavior, 66, 236–247. https://doi.org/10.1016/j.chb.2016.09.024
23. Kim, S. (2022). It is time that matters in crisis communication: The role of temporal distance and crisis threat appraisal. Public Relations Review, 48(2), 102155. https://doi.org/10.1016/j.pubrev.2022.102155
24. Krippendorff, K. (2008). Systematic and Random Disagreement and the Reliability of Nominal Data. Communication Methods and Measures, 2(4), 323–338. https://doi.org/10.1080/19312450802467134
25. Küçük, D., & Can, F. (2020). Stance Detection: A Survey. ACM Comput. Surv., 53(1). https://doi.org/10.1145/3369026
26. Landis, J. R., & Koch, G. G. (1977). The Measurement of Observer Agreement for Categorical Data. Biometrics, 33(1), 159–174. JSTOR. https://doi.org/10.2307/2529310
27. Laureate, C. D. P., Buntine, W., & Linger, H. (2023). A systematic review of the use of topic models for short text social media analysis. Artificial Intelligence Review, 56(12), 14223–14255. https://doi.org/10.1007/s10462-023-10471-x
28. Liyih, A., Anagaw, S., Yibeyin, M., & Tehone, Y. (2024). Sentiment analysis of the HamasIsrael war on YouTube comments using deep learning. Scientific Reports, 14(1), 13647. https://doi.org/10.1038/s41598-024-63367-3
29. López-Rabadán, P. (2022). Framing Studies Evolution in the Social Media Era. Digital Advancement and Reorientation of the Research Agenda. Social Sciences, 11(1), 9. https://doi.org/10.3390/socsci11010009
30. Madatov, K., Sattarova, S., & Vičič, J. (2025). TF-IDF-Based Classification of Uzbek Educational Texts. Applied Sciences, 15(19), 10808. https://doi.org/10.3390/app151910808
31. McHugh, M. L. (2012). Interrater reliability: The kappa statistic. Biochem Med (Zagreb), 22(3), 276–282. https://doi.org/10.11613/BM.2012.031
32. Mu, Y., Jin, M., Bontcheva, K., & Song, X. (2024). Examining Temporalities on Stance Detection towards COVID-19 Vaccination. In N. Calzolari, M.-Y. Kan, V. Hoste, A. Lenci, S. Sakti, & N. Xue (Eds.), Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024) (pp. 6732–6738). ELRA and ICCL. https://aclanthology.org/2024.lrec-main.594/
33. Muthusami, R., Saritha, K., Rao, K. S., Sugapriya, P., & Saveetha, G. (2025). Interpretable stance detection in social media via topic-guided transformers. Discover Artificial Intelligence, 5(1), 355. https://doi.org/10.1007/s44163-025-00635-9
34. Obadimu, A., Khaund, T., Mead, E., Marcoux, T., & Agarwal, N. (2021). Developing a socio-computational approach to examine toxicity propagation and regulation in COVID-19 discourse on YouTube. Information Processing & Management, 58(5), 102660. https://doi.org/10.1016/j.ipm.2021.102660
35. Park, C. S., & Molla, M. A. M. (2025). Collective Emotions on Social Media: An Examination of Collective Emotions in YouTube Comment Sections Below New Videos and Their Political Consequences. Howard Journal of Communications, 1–22. https://doi.org/10.1080/10646175.2025.2587082
36. Poell, T. (2020). Social media, temporality, and the legitimacy of protest. Social Movement Studies, 19(5–6), 609–624. https://doi.org/10.1080/14742837.2019.1605287
37. Reimers, N., & Gurevych, I. (2019). Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. arXiv. https://arxiv.org/abs/1908.10084
38. Silla, C. N., & Freitas, A. A. (2011). A survey of hierarchical classification across different application domains. Data Mining and Knowledge Discovery, 22(1), 31–72. https://doi.org/10.1007/s10618-010-0175-9
39. Sobhani, P., Inkpen, D., & Zhu, X. (2017). A Dataset for Multi-Target Stance Detection. In M. Lapata, P. Blunsom, & A. Koller (Eds.), Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers (pp. 551–557). Association for Computational Linguistics. https://aclanthology.org/E17-2088/
40. Substance Abuse and Mental Health Services Administration. (2010). Measures of Agreement and Approaches to Modeling. In Reliability of Key Measures in the National Survey on Drug Use and Health (Chapter 4). Substance Abuse and Mental Health Services Administration (US). https://www.ncbi.nlm. nih.gov/books/NBK519796/
41. Thomas, C., & Zanotti, J. (2025). Israel-Iran Conflict, U.S. Strikes, and Ceasefire (CRS In Focus). Congressional Research Service. https://www.congress.gov/crsproduct/IF13032
42. van der Breggen, M. M., Gonçalves, J., & Boeren, D. (2025). Polarization by recommendation: Analyzing YouTube’s polarization dynamics around Dutch political parties. Journal of Information Technology & Politics, 1–15. https://doi.org/10.1080/19331681.2025.2544653
43. Walter, D., & Ophir, Y. (2019). News Frame Analysis: An Inductive Mixed-method Computational Approach. Communication Methods and Measures, 13(4), 248–266. https://doi.org/10.1080/19312458.2019.1639145
44. Zhang, W., Yoshida, T., & Tang, X. (2008). Text classification based on multi-word with support vector machine. Knowledge-Based Systems, 21(8), 879–886. https://doi.org/10.1016/j.knosys.2008.03.044