Loading...
|
Please use this identifier to cite or link to this item:
https://nccur.lib.nccu.edu.tw/handle/140.119/155928
|
| Title: | 應用大型語言模型於公共政策網路輿情分析--以自動駕駛車議題為例
Applying Large Language Models to Public Policy Online Opinion Analysis: A Case Study on Autonomous Vehicles |
| Authors: | 陳智彬
Chen, Zhi-Bin |
| Contributors: | 蕭乃沂
Hsiao, Nai-Yi
陳智彬
Chen, Zhi-Bin |
| Keywords: | 大型語言模型
自駕車
公共政策
網路輿情分析
ChatGPT
TAIDE
Large language models
Autonomous vehicles
Public policy
Online sentiment analysis
ChatGPT
TAIDE |
| Date: | 2025 |
| Issue Date: | 2025-03-03 13:49:22 (UTC+8) |
| Abstract: | 本研究探討大型語言模型在公共政策網路輿情分析中的應用，選擇自動駕駛車（自駕車）議題作為案例，嘗試應用大型語言模型與輿情分析方法，以提升民意分析的效率並降低操作門檻。研究主要涵蓋網路輿情數據的收集與分析、大型語言模型的應用實驗，以及兩者的綜合比較。
研究通過分析2022年與2024年的網路輿情數據，揭示了台灣民眾對自駕車的意見多集中於技術進步與未來應用的討論，但也伴隨著對安全性和事故責任的高度關注。事件驅動特性在輿情中尤為顯著，例如技術發布或交通事故會大幅提升相關討論聲量。正向討論多與技術創新和便利性有關，而負向討論則聚焦於風險與適用性。這些觀察與過往文獻的靜態結果形成補充，展現了輿情的動態性。
應用ChatGPT與TAIDE進行的大型語言模型實驗顯示，模型在相關性與立場判斷上達到了接近人工標註的表現，特別是在經過微調（Fine-Tuning）後，相關性判斷準確率達88.78%，立場判斷準確率達79.56%，關鍵句提取的ROUGE指標也顯示良好的準確性。ChatGPT在性能與穩定性方面優於TAIDE，但TAIDE因其免費及本地化優勢，對強調數據隱私和經費考量的狀況下更具吸引力。
研究發現大型語言模型在輿情分析中的應用潛力巨大，能顯著降低人工標註的工作量，並在長期議題的追蹤分析中提供穩定表現。然而，模型在處理短文本、多義性語句以及避免AI幻覺時仍面臨挑戰。未來研究可以進一步優化提示詞設計、提升模型訓練數據品質，並擴展應用範疇，例如針對核能或死刑等其他公共政策議題進行長期監測分析。
整體而言，本研究證實了大型語言模型結合人工分析的可能性，展現了在公共政策輿情分析中的應用價值。未來透過更深度的技術整合與模型改進，將有助於實現更高效且全面的公共政策民意分析，為政策制定者提供更加即時且精確的參考依據。
This study explores the application of large language models (LLMs) in online public opinion analysis for public policy, using autonomous vehicles (AVs) as a case study. By leveraging LLMs and sentiment analysis methods, the research aims to enhance the efficiency of public opinion analysis while lowering operational barriers. The study encompasses the collection and analysis of online public opinion data, experimental applications of LLMs, and a comparative synthesis of the results.
Through an analysis of public opinion data from 2022 and 2024, the study reveals that discussions among the Taiwanese public about AVs are predominantly centered on technological advancements and future applications. However, concerns about safety and accident liability also feature prominently. The event-driven nature of public opinion is evident, with significant surges in discussion volume following technology launches or traffic incidents. Positive discussions often emphasize innovation and convenience, while negative opinions focus on risks and applicability. These findings supplement previous static literature by highlighting the dynamic nature of public sentiment.
Experimental applications using ChatGPT and TAIDE demonstrate that LLMs achieved performance levels close to human annotation in relevance and stance detection. After fine-tuning, the relevance detection accuracy reached 88.78%, stance detection accuracy reached 79.56%, and the ROUGE metrics for key sentence extraction showed strong precision. ChatGPT outperformed TAIDE in terms of performance and stability, but TAIDE’s cost-free and localized features make it attractive for scenarios prioritizing data privacy and budget considerations.
The study highlights the significant potential of LLMs in public opinion analysis, substantially reducing the workload of manual annotation and providing stable performance for long-term issue tracking. However, challenges remain in handling short texts, ambiguous statements, and mitigating AI hallucinations. Future research could further optimize prompt design, improve the quality of training data, and expand application domains, such as long-term monitoring of other public policy issues like nuclear energy or capital punishment.
In summary, this study confirms the feasibility of combining LLMs with manual analysis, showcasing their value in public policy sentiment analysis. Deeper technical integration and model improvements in the future can enhance the efficiency and comprehensiveness of public opinion analysis, offering policymakers more timely and accurate references. |
| Reference: | 中央研究院（2018）。斷開中文的鎖鍊！自然語言處理 (NLP)。7月3日。https://research.sinica.edu.tw/nlp-natural-language-processing-chinese-knowledge-information/
丘昌泰（2010）。公共政策。巨流圖書。
古倫維、陳信希（2019）。中文意見分析之概況、技術與應用。計算語言學通訊，20(5)，5-20。
甘偵蓉（2024）。AI開發過程的倫理權衡：自駕車決策案例研究。歐美研究，54(1)，1-67。
交通部（2024）。道安統計。2024年6月。https://stat.motc.gov.tw/mocdb/stmain.jsp?sys=100
吳長融（2020）。自然語言處理技術應用於中文網路新聞議題立場分析。〔未出版之碩士論文〕。中原大學。
吳奕靖（2023）。網路輿情與危機傳播研究-以高雄市政府處理新冠肺炎為例。〔未出版之碩士論文〕。義守大學。
吳泰毅、鄧玉羚（2023）。初探台灣民眾對人工智慧產品與服務之採用經驗與信任感。資訊社會研究，(44)，97-127。
呂建億（2015）。民眾對政府輿情分析方法之信任研究－民意調查與網路輿情分析的比較。〔未出版之碩士論文〕。國立政治大學。
巫家宇（2021）。新冠肺炎下的台灣網路輿情分析。〔未出版之博士論文〕。元智大學。
林文涵（2015）。網路輿情分析在公共政策的應用與影響。〔未出版之碩士論文〕。國立政治大學。
林佳蒼（2020）。多向注意力機制於翻譯任務改進之研究。〔未出版之碩士論文〕。國立中央大學。
邱凱隆（2023）。基於BERT與Ngram之混合模型於診斷證明書光學字元辨識後處理。〔未出版之碩士論文〕。元智大學。
財團法人國家實驗研究院（2024）。TAIDE - 推動臺灣可信任生成式AI發展計畫。https://taide.tw。
張倩瑜（2016）。食品安全風險議題研究：專家與民眾認知之比較。〔未出版之碩士論文〕。國立臺北大學。
張易筠（2022）。應用bert語言模型於顧客評論之多面向情緒分析。［未出版之碩士論文］。國立高雄科技大學資訊管理系。
莊文忠（2018）。循證的政策制定與資料分析：挑戰與前瞻。文官制度季刊，10(2)， 1-20。
莊鎮豪（2020）。深度學習與動態資料技術應用於語句反諷之分析。［未出版之碩士論文］。國立臺灣科技大學電機工程系。
郭毓倫（2021）。大數據視角下的公共政策－網路輿情分析方法之應用與發展。中國地方自治，74(9)，3-35。
陳鈺妏、林雅潔（2024）。全球自駕車發展現況與未來趨勢。財團法人車輛測試研究中心。https://www.artc.org.tw/tw/knowledge/articles/13760
陳品皓（2016）。網路使用行為對於台灣民眾政治參與的影響之初探研究。復興崗學報，(108)，95-120。
陳韋銘（2016）。使用輔助向量的雙邊特徵分群以改善中文新聞的立場偵測分類 ［未出版之碩士論文］。國立臺灣大學。
陳敦源、朱斌妤、蕭乃沂、黃東益、廖洲棚、曾憲立（2020）。政府數位轉型：一本必讀的入門書。五南。
陳群叡（2023）。政治輿論的迷因工程之論述研究：以網軍粉絲專頁2022年九合一選舉圖片貼文為例。［未出版之碩士論文］。國立臺灣師範大學。
陳威達（2020）。應用機器學習演算法進行文本情感分析之研究。［未出版之碩士論文］。德明財經科技大學資訊管理系。
黃丰嘉（2021）。基於自然語言處理技術整合維基百科 (Wiki) 之圖書館參考諮詢機器人建置與使用評估。［未出版之碩士論文］。輔仁大學。
黃東益（2014）。放射性廢棄物最終處置民眾關心議題蒐集與分析（編號：1032001INER047）。行政院原子能委員會委託研究案。
黃彪文、殷美香（2014）。從常人理論看專家與公眾對健康風險的認知差異。載於世新大學（主編），科學傳播論文集6（頁273-288）。世新大學。
黃瀚萱（2020）。政府治理智慧化：自然語言處理的可能性。國土及公共治理季刊，8(3)，30-37。
楊盛旺（2024）。自駕車法制初探-以道路交通相關法規為主（編號：1699）。立法院法制局專題研究報告。
廖洲棚、陳敦源、蕭乃沂、廖興中（2014）。運用巨量資料實踐良善治理：網路民意導入政府決策分析之可行性研究（編號：RDEC-MIS-102-003）。國家發展委員會。
劉芃葦（2016）。網路巨量時代下輿情意向之探究： 以我國自由經濟示範區政策為例。［未出版之碩士論文］。國立政治大學。
蕭乃沂（2023）。公共政策管理與決策導向的科技治理模式-以我國核能與人工智慧議題為個案（編號：MOST109-2511-H004-008-MY2）。行政院國家科學及技術委員會。
蕭乃沂、黃東益（2015）。 104年度網路輿情蒐集、分析及運用計畫結案報告。教育部青年發展署。
蕭乃沂、廖洲棚、陳敦源（2014）。政府應用巨量資料精進公共服務與政策分析之可行性研究（NDC-MIS-103-003）。行政院國家發展委員會。
賴思宇（2022）。以BERT模型為基礎之情緒分析研究-以Amazon評論為例。［未出版之碩士論文］。國立臺北大學。
Adams-Cohen, N. (2020). Policy Change and Public Opinion: Measuring Shifting Political Sentiment With Social Media Data. American Politics Research, 48, 612 - 621.
Adekanye, O. A. M. (2024). LLM-Powered Synthetic Environments for Self-Driving Scenarios. Proceedings of the AAAI Conference on Artificial Intelligence, 38(21), 23721-23723.
Anstead, N., & O'Loughlin, B. (2015). Social Media Analysis and Public Opinion: The 2010 UK General Election. Journal of Computer-Mediated Communication, 20(2), 204-220.
Barberá, P., Casas, A., Nagler, J., Egan, P., Bonneau, R., Jost, J. T., & Tucker, J. A. (2019). Who Leads? Who Follows? Measuring Issue Attention and Agenda Setting by Legislators and the Mass Public Using Social Media Data. The American Political Science Review,113, 883 - 901.
Berelson, B. (1952). Content analysis in communication research. Free Press.
Brown, T.B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D.M., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, A., Sutskever, I., & Amodei, D. (2020). Language Models are Few-Shot Learners. ArXiv, abs/2005.14165.
Bruno, A., Mazzeo, P.L., Chetouani, A., Tliba, M., & Kerkouri, M.A. (2023). Insights into Classifying and Mitigating LLMs' Hallucinations. AIxPAC@AI*IA.
Chen, X., Zeng, H., Xu, H.L., & Di, X. (2021). Sentiment Analysis of Autonomous Vehicles After Extreme Events Using Social Media Data. 2021 IEEE International Intelligent Transportation Systems Conference (ITSC), 1211-1216.
Cho, M., Schweickart, T., & Haase, A. (2014). Public engagement with nonprofit organizations on Facebook. Public Relations Review, 40(3), 565-567.
Cui, C., Ma, Y., Cao, X., Ye, W., & Wang, Z. (2023). Drive as You Speak: Enabling Human-Like Interaction with Large Language Models in Autonomous Vehicles. 2024 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW), 902-909.
Cui, C., Ma, Y., Cao, X., Ye, W., Zhou, Y., Liang, K., Chen, J., Lu, J., Yang, Z., Liao, K., Gao, T., Li, E., Tang, K., Cao, Z., Zhou, T., Liu, A., Yan, X., Mei, S., Cao, J., Wang, Z., & Zheng, C. (2023). A Survey on Multimodal Large Language Models for Autonomous Driving. 2024 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW), 958-979.
Damerau, F. J. (1971). Markov Models and Linguistic Theory: An Experimental Study of a Model for English. Mouton.
Davis, M. A., Zheng, K., Liu, Y., & Levy, H. (2017). Public Response to Obamacare on Twitter. Journal of Medical Internet Research, 19.
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. North American Chapter of the Association for Computational Linguistics.
dos Santos, F. L. M., Duboz, A., Grosso, M., Raposo, M. A., Krause, J., Mourtzouchou, A., Balahur, A., & Ciuffo, B. (2022). An acceptance divergence? Media, citizens and policy perspectives on autonomous cars in the European Union. Transportation Research Part A: Policy and Practice, 158, 224-238.
Dutta, A., Das, S. (2021). Tweets About Self-Driving Cars: Deep Sentiment Analysis Using Long Short-Term Memory Network (LSTM). In: Gupta, D., Khanna, A., Bhattacharyya, S., Hassanien, A.E., Anand, S., Jaiswal, A. (eds) International Conference on Innovative Computing and Communications. Advances in Intelligent Systems and Computing, vol 1165. Springer.
Hu, J.E., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., & Chen, W. (2021). LoRA: Low-Rank Adaptation of Large Language Models. ArXiv, abs/2106.09685.
Funkhouser, K. (2013). Paving the road ahead: autonomous vehicles, products liability, and the need for new approach. Utah Law Review, 2013(1), 437-vi.
Gao, Y., Tong, W., Wu, E. Q., Chen, W., Zhu, G., & Wang, F. Y. (2023). Chat With ChatGPT on Interactive Engines for Intelligent Driving. IEEE Transactions on Intelligent Vehicles, 8(3), 2034-2036.
Ginsberg, J., Mohebbi, M. H., Patel, R. S., Brammer, L., Smolinski, M. S., & Brilliant, L. (2009). Detecting influenza epidemics using search engine query data. Nature, 457(7232), 1012-1014.
Gintova, M. (2019). Understanding government social media users: an analysis of interactions on Immigration, Refugees and Citizenship Canada Twitter and Facebook. Government Information Quarterly, 36(4), 101388.
Kacperski, C., Kutzner, F., & Vogel, T. (2021). Consequences of autonomous vehicles: Ambivalent expectations and their impact on acceptance. Transportation Research Part F: Traffic Psychology and Behaviour, 81, 282-294.
Kim, D. S., & Kim, J. W. (2014). Public Opinion Sensing and Trend Analysis on Social Media: A Study on Nuclear Power on Twitter. International Journal of Multimedia and Ubiquitous Engineering, 9(11), 373-384.
Kojima, T., Gu, S. S., Reid, M., Matsuo, Y., & Iwasawa, Y. (2024). Large language models are zero-shot reasoners Proceedings of the 36th International Conference on Neural Information Processing Systems, New Orleans, LA, USA.
Lampos, V., Majumder, M., Yom-Tov, E., Edelstein, M., Moura, S., Hamada, Y., Rangaka, M., McKendry, R., & Cox, I. (2021). Tracking COVID-19 using online search. npj Digital Medicine, 4(17).
Lesteven, G., & Thébert, M. (2022). Who cares about AVs? Insights from French media discourse on Twitter and in the press. Case Studies on Transport Policy, 10(2), 1078-1089.
LeValley, D. (2013). Autonomous Vehicle Liability—Application of Common Carrier Liability.
Li, T., Lin, L., Choi, M., Fu, K., Gong, S., & Wang, J. (2018). YouTube AV 50K: An Annotated Corpus for Comments in Autonomous Vehicles. 2018 International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP), 1-5.
Lin, C. (2004). ROUGE: A Package for Automatic Evaluation of Summaries. Annual Meeting of the Association for Computational Linguistics.
Lippmann, W. (1922). Public opinion. Harcourt.
Litman (2017) Autonomous Vehicle Implementation Predictions. Implications for Transport Planning. 8 September 2017. Victoria Transport Policy Institute.
Ma, J., Li, J., Gao, W., Yang, Y., & Wong, K. (2023). Improving Rumor Detection by Promoting Information Campaigns With Transformer-Based Generative Adversarial Learning. IEEE Transactions on Knowledge and Data Engineering, 35, 2657-2670.
Ma, W., & Chen, K. (2003). Introduction to CKIP Chinese Word Segmentation System for the First International Chinese Word Segmentation Bakeoff. Workshop on Chinese Language Processing.
Nordberg, P., Kävrestad, J., & Nohlberg, M. (2020). Automatic Detection of Fake News. Proceedings of the 6th International Workshop on Socio-Technical Perspective in IS Development (STPIS 2020) : Virtual Conference in Grenoble, France, June 8-9, 2020, 168–179.
O'Connor, B., Balasubramanyan, R., Routledge, B., & Smith, N. (2010, May). From tweets to polls: Linking text sentiment to public opinion time series. In Proceedings of the international AAAI conference on web and social media, 4(1), 122-129.
OpenAI. (2015). OpenAI官方網站. Retrieved May 30, 2024, from https://openai.com/
Pang, B., & Lee, L. (2008). Opinion Mining and Sentiment Analysis. Found. Trends in Information Retrieval, 2(1–2), 1–135.
Papineni, K., Roukos, S., Ward, T., & Zhu, W.-J. (2002). Papineni, K., Roukos, S., Ward, T., & Zhu, W. (2002). Bleu: a Method for Automatic Evaluation of Machine Translation. Annual Meeting of the Association for Computational Linguistics.
Ratnayake, H., & Wang, C. (2024, 2024//). A Prompting Framework to Enhance Language Model Output. AI 2023: Advances in Artificial Intelligence, Singapore.
Raza, M., Awais, M., Ali, K., Aslam, N., Paranthaman, V. V., Imran, M., & Ali, F. (2020). Establishing effective communications in disaster affected areas and artificial intelligence based detection using social media platform. Future Generation Computer Systems, 112, 1057-1069.
Reisach, U. (2021). The responsibility of social media in times of societal and political manipulation. European Journal of Operational Research, 291(3), 906-917.
Reveilhac, M., & Eisner, L. (2022). Political Polarisation on Gender Equality: The Case of the Swiss Women’s Strike on Twitter. Statistics, Politics and Policy, 13, 255 - 278.
Reveilhac, M., Steinmetz, S., & Morselli, D. (2021). A systematic literature review of how and whether social media data can complement traditional survey data to study public opinion. Multimedia Tools and Applications, 2022(81), 10107-10142.
SAE. (2018). SAE International Releases Updated Visual Chart for Its “Levels of Driving Automation” Standard for Self-Driving Vehicles. Retrieved May 30 from https://www.sae.org/news/press-room/2018/12/sae-international-releases-updated-visual-chart-for-its-%E2%80%9Clevels-of-driving-automation%E2%80%9D-standard-for-self-driving-vehicles
Schneble, C. O., & Shaw, D. M. (2021). Driver’s views on driverless vehicles: Public perspectives on defining and using autonomous cars. Transportation Research Interdisciplinary Perspectives, 11 (7729), 100446.
Suen, C.Y. (1979). n-Gram Statistics for Natural Language Understanding and Text Processing. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-1, 164-172.
Soleimani, A., Monz, C., & Worring, M. (2019). BERT for Evidence Retrieval and Claim Verification. Advances in Information Retrieval, 12036, 359 - 366.
Strapparava, C., & Valitutti, A. (2004). WordNet Affect: an Affective Extension of WordNet. International Conference on Language Resources and Evaluation.
Tran, V., & Matsui, T. (2023, October). Public Opinion Mining Using Large Language Models on COVID-19 Related Tweets. In 2023 15th International Conference on Knowledge and Systems Engineering (KSE) (pp. 1-6). IEEE.
Wang, Z., Xie, Q., Ding, Z., Feng, Y., & Xia, R. (2023). Is ChatGPT a Good Sentiment Analyzer? A Preliminary Study. ArXiv, abs/2304.04339.
Wei, J., Bosma, M., Zhao, V., Guu, K., Yu, A. W., Lester, B., Du, N., Dai, A. M., & Le, Q. V. (2021). Finetuned Language Models Are Zero-Shot Learners. ArXiv, abs/2109.01652.
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E. H., Le, Q. V., & Zhou, D. (2024). Chain-of-thought prompting elicits reasoning in large language models. Proceedings of the 36th International Conference on Neural Information Processing Systems, New Orleans, LA, USA.
Zhang, L., Wang, S., & Liu, B. (2018). Deep learning for sentiment analysis: A survey. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 8(4), e1253.
Zhang, W., Deng, Y., Liu, B.-Q., Pan, S. J., & Bing, L. (2023). Sentiment Analysis in the Era of Large Language Models: A Reality Check. ArXiv, abs/2305.15005.
Zheng, Y., Zhang, R., Zhang, J., Ye, Y., Luo, Z., & Ma, Y. (2024). LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models. ArXiv, abs/2403.13372. |
| Description: | 碩士
國立政治大學
公共行政學系
110256029 |
| Source URI: | http://thesis.lib.nccu.edu.tw/record/#G0110256029 |
| Data Type: | thesis |
| Appears in Collections: | [公共行政學系] 學位論文
|
Files in This Item:
| File |
Description |
Size | Format | |
|---|
| 602901.pdf | | 6395Kb | Adobe PDF | 0 | View/Open |
|
All items in 政大典藏 are protected by copyright, with all rights reserved.
|
