全球氣候變遷已成為當代重要議題，碳排放作為氣候變遷的核心因素，受到各界的高度關注。企業社會責任揭露文件也逐漸普及，高層管理者在文件中的語氣隱含的資訊應被重視。然而專注在碳排放議題上的文本分析因缺乏專屬字典，所以較少有相關研究深入探討。本研究旨在填補此空白，期望找出有效建立碳排放字典的方法並嘗試建立碳排放字典，深化文本分析工具的應用。本研究除了以次級資料分析法，還嘗試了以ChatGPT-4o產生字詞，最後結合專家意見法，以專家之開放式問卷、半結構式訪談逐步提升字典的專業性、全面性與前後文關聯性。本研究蒐集企業永續報告書、財務報告做次級資料分析，並以道瓊工業指數與標準普爾100指數中37家美國上市企業2020年至2022年永續報告書中高層管理者的言論為研究文本，結合本研究構建的碳排放字典、BERT及ChatGPT-4o進行情感分析。研究結果顯示，採用第三階段字典的分析模型與專家情感評分間的相關係數最高；此外，ChatGPT-4o在情感分析與專家情感評分相關性檢定方面，因其卓越的長文本處理能力，在揭示文本隱含碳排放情感傾向方面展現顯著優勢。在碳排放字典的應用面，以簡單線性迴歸可以得出高層管理者的負向情感與財務槓桿率具有顯著關聯性。本研究為碳排放情感分析提供新的方法與視角，以協助利益關係人了解企業在其自主揭露文件中對碳排放議題的語言策略，並為未來相關研究提供實質性的參考依據。

Global climate change has emerged as a pressing issue of our time, with carbon emissions at its core garnering significant attention from various sectors. Corporate social responsibility disclosures are becoming increasingly common, and the tone conveyed by top management in these documents carries valuable information that deserves greater attention. However, research that focuses on analyzing carbon emissions in textual data remains limited, partly due to the absence of a specialized carbon emissions dictionary. This study aims to identify effective methods for constructing a carbon emissions dictionary and attempts to develop such a dictionary to enhance the application of text analysis tools. In addition to employing secondary data analysis, the research also explores the use of ChatGPT-4o to generate relevant terms. Finally, expert opinion methods are integrated, utilizing open-ended questionnaires and semi-structured interviews with experts to progressively enhance the dictionary's professionalism, comprehensiveness, and contextual relevance. We collected sustainability and financial reports from corporations listed in the Dow Jones Industrial Average and the S&P 100 indexes, selecting 37 U.S.-listed firms and their sustainability reports—specifically, the statements of senior executives from 2020 to 2022—as our primary textual corpus. By leveraging a newly constructed carbon emissions dictionary, along with BERT and ChatGPT-4o for sentiment analysis, we conducted an in-depth examination of these statements.
The results indicate that the analytical model employing the third version of the dictionary showed the highest correlation with expert sentiment assessments. Moreover, ChatGPT-4o demonstrates significant advantages in sentiment analysis and correlation assessment with expert sentiment ratings, owing to its exceptional capability in processing long texts. This enables it to effectively uncover implicit sentiment tendencies related to carbon emissions within textual data. In terms of applying the carbon emissions dictionary, a simple linear regression revealed that negative sentiments expressed by top management have a significant relationship with firms’ financial leverage ratios. This study provides a novel methodology and perspective for carbon emissions sentiment analysis, assisting stakeholders in understanding the linguistic strategies that corporations adopt in their voluntary disclosure documents regarding carbon emissions, and offering substantive insights for future related research.

主目錄
第一章 緒論	1
第一節	研究背景與動機	1
第二節	研究目的	3
第二章 文獻探討	5
第一節	字典法作為情感分析工具	5
第二節	BERT作為情感分析工具	7
第三節	LLM作為情感分析工具	8
第四節	永續議題的情感分析	9
第五節	情感分析與財務槓桿率	11
第六節	文獻總結與研究缺口	14
第三章 研究方法與研究設計	18
第一節	研究流程	18
第二節	研究設計	21
第四章 研究內容與分析	31
第一節	企業永續報告書整理	31
第二節	各階段碳排放字典	31
第三節	情感分析組合之評分者間信度檢驗	42
第四節	敘述性統計	46
第五節	碳排放情感對於財務槓桿率的關聯性	46
第五章 研究討論	49
第一節	碳排放字典	49
第二節	碳排放情感分析流程	51
第三節	高層管理者碳排放情感與財務槓桿率的關聯性	56
第六章 結論與建議	59
第一節	結論	59
第二節	管理意涵	64
第三節	研究限制與建議	65
參考文獻	68
附錄A.本研究樣本企業列表	78
附錄B.建立碳字典第二階段-開放式問卷說明	79
附錄C.半結構式訪談問卷	80
附錄D.專家碳排放情感評分階段-開放式問卷說明	81

圖目錄
圖3-1研究流程	20
圖3-2研究架構	21
圖3-3專家問卷、訪談與詞彙涵蓋率驗證流程圖	26
圖3-4研究假設	29
圖4-1第一階段擴展字詞生成字典之指令(Prompt)	32
圖4-2第三階段字典字詞相關性之指令(Prompt)	41
圖4-3 ChatGPT-4o情感分析之指令(Prompt)	45

表目錄
表2-1 研究缺口摘要	16
表4-1第一階段碳排放相關詞彙	32
表4-2詞彙涵蓋率相關文獻整理	35
表4-3專家彙整表	37
表4-4第二階段新增之碳排放相關詞彙	38
表4-5專家建議增加詞彙說明彙整	38
表4-6第三階段碳排放字典詞彙	41
表4-7各項情感定義	42
表4-8專家評分正向情感相關係數	43
表4-9專家評分中立情感相關係數	43
表4-10專家評分負向情感相關係數	44
表4-11專家評分綜合情感相關係數	44
表4-12 BERT、ChatGPT-4o與專家情感分數之相關係數	45
表4-13永續報告中高層管理者碳排放情感敘述性統計	46
表4-14高層管理者碳排放情感與財務槓桿率之簡單線性迴歸	47
表4-15假設檢驗結果	47
表5-1 BERT與ChatGPT-4o情感分析比較	53
表5-2 BERT、ChatGPT-4o、RAG與專家情感分數之相關係數	56

王雅詩(2017)。基於詞性組合的意見字典擴增方法之研究。淡江大學，碩士論文。
白明弘、吳鑑城、簡盈妮、黃淑齡、林慶隆(2016)。基於詞語分佈均勻度的核心詞彙選擇。中文計算語言學期刊，21(2)，1-17頁。
何莉芸、李佳玲(2017)。企業永續發展與社會責任報告之趨勢與實務。中華會計學刊，第12特刊卷，1-24頁。
李嘉琪(2024)。從永續報告書比較台灣與亞洲其他電信公司之永續作為。國立中山大學，碩士論文。
高韶宇(2024)。從TCFD建議分析臺灣銀行業之氣候財務揭露。國立中山大學，碩士論文。
陳淑妙(2023)。走在不斷精進的道路上：一位視障按摩師進修物理治療的歷程。台南大學，碩士論文。
萬文隆(2004)。深度訪談在質性研究中的應用。生活科技教育月刊。
楊懿麗(2006)。國內各級英語教學的詞彙量問題。國立編譯館館刊，34(3)，35-44頁。
趙文裕(2022)。港埠倉儲業者導入溫室氣體盤查個案分析。逢甲大學，碩士論文。
謝靜婷(2008)。半自動建立中文Wordnet之研究。國立清華大學，碩士論文。
Abeydeera, L. H. U. W., Mesthrige, J. W., Samarasinghalage, T. I. (2019). Global research on carbon emissions: A scientometric review. Sustainability, 11(14).
Amel-Zadeh, A., Chen, M., Mussalli, G., & Weinberg, M. (2021). NLP for SDGs: measuring corporate alignment with the sustainable development goals. Columbia Business School Research Paper. 
Amernic, J., Craig, R., & Tourish, D. (2010). Measuring and assessing tone at the top using annual report CEO letters. The Institute of Chartered Accountants of Scotland. 
Araci, D. (2019). FinBERT: Financial Sentiment Analysis with Pre-trained Language Models. arXiv preprint arXiv:1908.10063.
Baier, P., Berninger, M., & Kiesel, F. (2020). Environmental, social and governance reporting in annual reports: A textual analysis. Financial Markets, Institutions & Instruments, 29(3), 93-118. 
Ben-David, I., Graham, J. R., & Harvey, C. R. (2013). Managerial miscalibration. The Quarterly Journal of Economics, 128(4), 1547-1584.
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March). On the dangers of stochastic parrots: Can language models be too big?. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610-623. 
Boiral, O., & Heras-Saizarbitoria, I. (2020). Sustainability reporting assurance: creating stakeholder accountability through hyperreality? Journal of Cleaner Production, 243, 1-17.
Botosan, C. A. (1997). Disclosure level and the cost of equity capital. Accounting Review, 323-349.
Bradburn, N. M., Sudman, S., & Wansink, B. (2004). Asking questions: The definitive guide to questionnaire design: For market research, political polls, and social and health questionnaires. San Francisco, CA: Jossey-Bass.
Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., & Askell, A. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877-1901.
Cao, Y., Myers, L. A., & Omer, T. C. (2012). Does company reputation matter for financial reporting quality? Evidence from restatements. Contemporary Accounting Research, 29(3), 956-990.
Chatfield, S. L. (2020). Recommendations for secondary analysis of qualitative data. The Qualitative Report, 25(3), 833-842.
Cheng, B., Ioannou, I., & Serafeim, G. (2014). Corporate social responsibility and access to finance. Strategic Management Journal, 35(1), 1-23.
Cho, C. H., Michelon, G., & Patten, D. M. (2012). Impression management in sustainability reports: An empirical investigation of the use of graphs. Accounting and the Public Interest, 12(1), 16-37.
Christensen, H. B., Hail, L., & Leuz, C. (2021). Mandatory CSR and sustainability reporting: Economic analysis and literature review. Review of Accounting Studies, 26(3), 1176-1248. 
Cong, Y., Freedman, M., & Park, J. D. (2014). Tone at the top: CEO environmental rhetoric and environmental performance. Advances in Accounting, 30(2), 322-327. 
Cuconasu, F., Trappolini, G., Siciliano, F., Filice, S., Campagnano, C., Maarek, Y., Tonellotto, N., & Silvestri, F. (2024). The power of noise: Redefining retrieval for rag systems. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, 719-729.
Derrick, K. (2024). ESG Sentiment Analysis: comparing human and language model performance including GPT. arXiv preprint arXiv:2402.16650.
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. arXiv (Cornell University).
Dhaliwal, D. S., Li, O. Z., Tsang, A., & Yang, Y. G. (2011). Voluntary nonfinancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting. The Accounting Review, 86(1), 59-100. 
El Ghoul, S., Guedhami, O., Kwok, C. C., & Mishra, D. R. (2011). Does corporate social responsibility affect the cost of capital? Journal of Banking & Finance, 35(9), 2388-2406.
Flammer, C. (2015). Does corporate social responsibility lead to superior financial performance? A regression discontinuity approach. Management Science, 61(11), 2549-2568.
Górka, J., & Kuziak, K. (2022). Volatility modeling and dependence structure of ESG and conventional investments. Risks, 10(1), 20.
Goss, A., & Roberts, G. S. (2011). The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance, 35(7), 1794-1810.
Graham, J. R., Harvey, C. R., & Puri, M. (2013). Managerial attitudes and corporate actions. Journal of Financial Economics, 109(1), 103-121.
Harymawan, I., Nasih, M., Ratri, M. C., Soeprajitno, R. R. W. N., & Shafie, R. (2020). Sentiment analysis trend on sustainability reporting in Indonesia: Evidence from construction industry. Journal of Security and Sustainability Issues, 9(3).
Heaton, J. B. (2002). Managerial optimism and corporate finance. Financial Management-Tampa -, 31(2; SEAS SUM), 33-46.
Hu, M., & Liu, B. (2004). Mining and summarizing customer reviews. Proceedings of the Tenth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
Huang, A. H., Wang, H., & Yang, Y. (2023). FinBERT: A large language model for extracting information from financial text. Contemporary Accounting Research, 40(2), 806-841. 
Huang, F., Chen, M., & Liu, R. (2023). The nature of corporate social responsibility disclosure and investment efficiency: Evidence from China. Frontiers in Environmental Science, 11, 1028745. 
Hunton, J. E., Hoitash, R., & Thibodeau, J. C. (2011). Retracted: the relationship between perceived tone at the top and earnings quality. Contemporary Accounting Research, 28(4), 1190-1224. 
Hutto, C., & Gilbert, E. (2014). Vader: A parsimonious rule-based model for sentiment analysis of social media text. In Proceedings of the International AAAI Conference on Web and Social Media. 
International, K. (2022). Global Survey of Sustainability Reporting 2022. In: KPMG International.
Kang, J., & Kim, Y. H. (2014). The impact of media on corporate social responsibility. Available at SSRN 2287002. 
Kothari, S. P., Li, X., & Short, J. E. (2009). The effect of disclosures by management, analysts, and business press on cost of capital, return volatility, and analyst forecasts: A study using content analysis. The Accounting Review, 84(5), 1639-1670.
Li, F. (2010). Textual analysis of corporate disclosures: A survey of the literature. Journal of Accounting Literature, 29(1), 143-165.
Lins, K. V., Servaes, H., & Tamayo, A. (2017). Social capital, trust, and firm performance: The value of corporate social responsibility during the financial crisis. The Journal of Finance, 72(4), 1785-1824.
Liu, B. (2012). Sentiment Analysis and Opinion Mining. Springer Nature. 
Loughran, T., & McDonald, B. (2011). When is a liability not a liability? Textual analysis, dictionaries, and 10‐Ks. The Journal of Finance, 66(1), 35-65. 
Loughran, T., & McDonald, B. (2016). Textual analysis in accounting and finance: A survey. Journal of Accounting Research, 54(4), 1187-1230. 
Mahadeo, S. K. (2006). English language teaching in Mauritius: A need for clarity of vision regarding English language policy. The International Journal of Language, Society and Culture, 18, 18-12. 
Malmendier, U., & Tate, G. (2015). Behavioral CEOs: The role of managerial overconfidence. Journal of Economic Perspectives, 29(4), 37-60.
Malmendier, U., Tate, G., & Yan, J. (2011). Overconfidence and early‐life experiences: the effect of managerial traits on corporate financial policies. The Journal of Finance, 66(5), 1687-1733.
Michelon, G., Patten, D. M., & Romi, A. M. (2019). Creating legitimacy for sustainability assurance practices: Evidence from sustainability restatements. European Accounting Review, 28(2), 395-422.
Ng, A. C., & Rezaee, Z. (2015). Business sustainability performance and cost of equity capital. Journal of Corporate Finance, 34, 128-149.
Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations and Trends in Information Retrieval, 2(1–2), 1-135.
Pang, S., Nol, E., & Heng, K. (2024). ChatGPT-4o for English language teaching and learning: Features, applications, and future prospects. Available at SSRN 4837988.
Park, S., & Kim, Y. (2016). Building thesaurus lexicon using dictionary-based approach for sentiment classification. In Software Engineering Research, Management and Applications (SERA). In Proceedings of the 2016 IEEE 14th International Conference, 39-44.
Patelli, L., & Pedrini, M. (2015). Is tone at the top associated with financial reporting aggressiveness? Journal of Business Ethics, 126, 3-19. 
Qiu, Y., Shaukat, A., & Tharyan, R. (2016). Environmental and social disclosures: Link with corporate financial performance. The British Accounting Review, 48(1), 102-116.
Reja, U., Manfreda, K. L., Hlebec, V., & Vehovar, V. (2003). Open-ended vs. close-ended questions in web questionnaires. Developments in Applied Statistics, 19(1), 159-177.
Rizinski, M., Peshov, H., Mishev, K., Jovanovik, M., & Trajanov, D. (2024). Sentiment Analysis in Finance: From Transformers Back to eXplainable Lexicons (XLex). IEEE Access.
Rocca, L., Giacomini, D., & Zola, P. (2021). Environmental disclosure and sentiment analysis: state of the art and opportunities for public-sector organisations. Meditari Accountancy Research, 29(3), 617-646.
Ross, S. A., Jaffe, J., & Kakani, R. K. (2019). Corporate Finance, 10e. McGraw-Hill Education.
Ruggiano, N., & Perry, T. E. (2019). Conducting secondary analysis of qualitative data: Should we, can we, and how? Qualitative Social Work, 18(1), 81-97.
Sharfman, M. P., & Fernando, C. S. (2008). Environmental risk management and the cost of capital. Strategic Management Journal, 29(6), 569-592.
Sol, K., Heng, K., & Sok, S. (2024). Using AI in English language education: An exploration of Cambodian EFL university students’ experiences, perceptions, and attitudes. Perceptions, and Attitudes.
Stemler, S. E. (2019). A comparison of consensus, consistency, and measurement approaches to estimating interrater reliability. Practical Assessment, Research, and Evaluation, 9(1), 4.
Sun, C., Huang, L., & Qiu, X. (2019). Utilizing BERT for aspect-based sentiment analysis via constructing auxiliary sentence. arXiv preprint arXiv:1903.09588.
Taboada, M., Brooke, J., Tofiloski, M., Voll, K., & Stede, M. (2011). Lexicon-based methods for sentiment analysis. Computational Linguistics, 37(2), 267-307. 
Testarmata, S., Fortuna, F., & Ciaburri, M. (2018). The communication of corporate social responsibility practices through social media channels. Corporate Board Role Duties and Composition, 14(1), 34-49. 
Tetlock, P. C. (2007). Giving content to investor sentiment: The role of media in the stock market. The Journal of Finance, 62(3), 1139-1168.
Thadewald, T., & Büning, H. (2007). Jarque–Bera test and its competitors for testing normality–a power comparison. Journal of Applied Satistics, 34(1), 87-105.
Torroni, P., Almeida, V. d. R. S., & Ruberg, N. (2021). BERT goes sustainable: an NLP approach to ESG financing.
Wang, Z., Zhu, Y., & Zhang, Q. (2024). LLM for sentiment analysis in e-commerce: A deep dive into customer feedback. Applied Science and Engineering Journal for Advanced Research, 3(4), 8-13. 
Wickham, R. J. (2019). Secondary analysis research. Journal of the Advanced Practitioner in Oncology, 10(4), 395.
Winata, G. I., Madotto, A., Lin, Z., Liu, R., Yosinski, J., & Fung, P. (2021). Language models are few-shot multilingual learners. arXiv preprint arXiv:2109.07684. 
Wu, H. H., Tsai, A. C. R., Tsai, R. T. H., & Hsu, J. Y. J. (2013). Building a Graded Chinese Sentiment Dictionary Based on Commonsense Knowledge for Sentiment Analysis of Song Lyrics. Journal of Information Science & Engineering, 29(4), 647-662.
Wu, S., Zhang, H., & Wei, T. (2021). Corporate social responsibility disclosure, media reports, and enterprise innovation: evidence from Chinese listed companies. Sustainability, 13(15), 8466. 
Zhang, B., Yang, H., Zhou, T., Ali Babar, M., & Liu, X. Y. (2023). Enhancing financial sentiment analysis via retrieval augmented large language models. In Proceedings of the Fourth ACM International Conference on AI in Finance, 349-356.
Zhang, W., Deng, Y., Liu, B., Pan, S., & Bing, L. (2024). Sentiment Analysis in the Era of Large Language Models: A Reality Check. In Proceedings of the Findings of the Association for Computational Linguistics: NAACL 2024.