臺灣博碩士論文加值系統

近年來，深度學習在自然語言處理的領域吹起一陣旋風，分別在不同的任務中皆取得優良成果，包括機器翻譯與 文字探勘 。 而過往研究 CSR報告的論文大部分停留在採取在基礎的分析方式， 故本研究運用深度學習技術嘗試探討 CSR報告的資訊內涵， 直接 將 CSR報告內容轉換成更為直觀 的 情感分數。 希望能成為財務指標之外，進行企業分析的重要變數。
本研究以英國上市櫃公司的 CSR報告為樣本，透過深度學習萃取出的情緒分數，以迴歸模型為分析方法，探討 CSR報告情緒對企業財務績效的影響。實證結果顯示， CSR報告的情緒與企業的財務市場績效產生顯著正相關。

In recent years, deep learning has been widely discussed in the field of natural language processing, and has achieved excellent results in different tasks, including machine translation and text mining. Most of the research of CSR report are based on the basic analysis method. Therefore, this study apply the deep learning method to try to get the information content of CSR report, and directly convert the CSR content into a more intuitive sentiment score. Hopefully it can be an important indicator of corporate analysis in addition to financial indicators.
In order to understand the influence of sentiment (which is predicted by the deep learning model) on the financial performance of firms, We collect CSR reports from the British firms which are on the London stock exchange as our samples. Using the linear regression model, the result showed that there is a strong positive relationship between the sentiment and the firm’s financial market performance.

摘要 i
Abstract ii
表目錄 v
圖目錄 vi
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 2
1.3 章節架構 3
第二章 文獻探討 5
2.1 CSR報告相關文獻 5
2.2 深度學習 (Deep Learning)於情感分析相關文獻 7
2.2.1 卷積神經網絡 (Convolutional Neural Network : CNN) 7
2.2.2 長短期記憶 (Long Short Term Memory Network : LSTM) 7
2.2.3 Gated Recurrent Unit (GRU): 8
2.2.4 集成式學習 (Ensemble learning) 8
2.3 神經網路的訓練方法 9
2.4 詞向量模型 10
2.5 情感分析相關論文 11
第三章 研究方法與資料 12
3.1 資料來源與樣本前處理 12
3.1.1 資料來源 12
3.1.2 樣本前處理 13
3.2 句子情緒數值分類 15
3.3 深度學習模型 19
3.3.1 句子詞向量構成 20
3.3.2 GRU-CNN模型 20
3.3.3 損失函數 (loss funtion) 23
3.3.4 模型參數初始化 23
3.3.5 超參數調整 24
3.3.6 模型比較驗證 24
3.4 預測情緒之衡量 25
3.5 迴歸模型設計 26
3.5.1 公司財務性績效之衡量 26
3.5.2 假說建立 27
3.5.3 第一組 Y變數 :Tobin’s Q(t+1)、 Tobin’s Q(t) 27
3.5.4 第二組 Y變數 :ROA(t+1)/ROE(t+1)與 ROA(t)/ROE(t) 29
第四章 實證結果 33
4.1 敘述性統計與相關係數矩陣 33
4.2 實證結果 36
4.2.1 假說 1與假說 2驗證 36
4.2.2 假說 3與假說 4驗證 39
4.2.3 延伸研究 40
第五章 結論與建議 41
5.1 研究結果 41
5.2 研究限制與建議 42
5.2.1 研究限制 42
5.2.2 研究建議 43
參考文獻 44
附錄 47

Akhtar, M. S., Kumar, A., Ghosal, D., Ekbal, A., & Bhattacharyya, P. (2017). A multilayer perceptron based ensemble technique for fine-grained financial sentiment analysis. Paper presented at the Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing.
Amini, M., Bienstock, C. C., & Narcum, J. A. (2018). Status of corporate sustainability: a content analysis of Fortune 500 companies. Business Strategy and the Environment, 27(8), 1450-1461.
Aureli, S. (2017). A comparison of content analysis usage and text mining in CSR corporate disclosure. International Journal of Digital Accounting Research.
Barontini, R., & Caprio, L. (2006). The effect of family control on firm value and performance: Evidence from continental Europe. European Financial Management, 12(5), 689-723.
Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words (ANEW): Instruction manual and affective ratings.
Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078.
Connolly, R. A., & Hirschey, M. (2005). Firm size and the effect of R&D on Tobin''s q. R&d Management, 35(2), 217-223.
Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., & Harshman, R. (1990). Indexing by latent semantic analysis. Journal of the American society for information science, 41(6), 391-407.
Deng, Y., Bao, F., Kong, Y., Ren, Z., & Dai, Q. (2016). Deep direct reinforcement learning for financial signal representation and trading. IEEE transactions on neural networks and learning systems, 28(3), 653-664.
Elliott, W. B., Grant, S. M., & Rennekamp, K. M. (2017). How disclosure features of corporate social responsibility reports interact with investor numeracy to influence investor judgments. Contemporary Accounting Research, 34(3), 1596-1621.
Glorot, X., & Bengio, Y. (2010). Understanding the difficulty of training deep feedforward neural networks. Paper presented at the Proceedings of the thirteenth international conference on artificial intelligence and statistics.
Greff, K., Srivastava, R. K., Koutník, J., Steunebrink, B. R., & Schmidhuber, J. (2016). LSTM: A search space odyssey. IEEE transactions on neural networks and learning systems, 28(10), 2222-2232.
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
Kim, Y. (2014). Convolutional neural networks for sentence classification. arXiv preprint arXiv:1408.5882.
Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.
Kiritchenko, S., Zhu, X., & Mohammad, S. M. (2014). Sentiment analysis of short informal texts. Journal of Artificial Intelligence Research, 50, 723-762.
Liao, P.-C., Xia, N.-N., Wu, C.-L., Zhang, X.-L., & Yeh, J.-L. (2017). Communicating the corporate social responsibility (CSR) of international contractors: Content analysis of CSR reporting. Journal of Cleaner Production, 156, 327-336.
Liu, S.-H., Chen, S.-Y., & Li, S.-T. (2017). Text-Mining Application on CSR Report Analytics: A Study of Petrochemical Industry. Paper presented at the 2017 6th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI).
Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013). Distributed representations of words and phrases and their compositionality. Paper presented at the Advances in neural information processing systems.
Muslu, V., Mutlu, S., Radhakrishnan, S., & Tsang, A. (2017). Corporate social responsibility report narratives and analyst forecast accuracy. Journal of Business Ethics, 1-24.
Pennington, J., Socher, R., & Manning, C. (2014). Glove: Global vectors for word representation. Paper presented at the Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP).
Polyak, B. T. (1964). Some methods of speeding up the convergence of iteration methods. USSR Computational Mathematics and Mathematical Physics, 4(5), 1-17.
Ravisankar, P., Ravi, V., Rao, G. R., & Bose, I. (2011). Detection of financial statement fraud and feature selection using data mining techniques. Decision Support Systems, 50(2), 491-500.
Russell, J. A. (1980). A circumplex model of affect. Journal of personality and social psychology, 39(6), 1161.
Servaes, H., & Tamayo, A. (2013). The impact of corporate social responsibility on firm value: The role of customer awareness. Management science, 59(5), 1045-1061.
Soo, C. K. (2018). Quantifying sentiment with news media across local housing markets. The Review of Financial Studies, 31(10), 3689-3719.
Steinert, L., & Herff, C. (2018). Predicting altcoin returns using social media. PloS one, 13(12), e0208119.
Tai, K. S., Socher, R., & Manning, C. D. (2015). Improved semantic representations from tree-structured long short-term memory networks. arXiv preprint arXiv:1503.00075.
Tang, D., Qin, B., & Liu, T. (2015). Document modeling with gated recurrent neural network for sentiment classification. Paper presented at the Proceedings of the 2015 conference on empirical methods in natural language processing.
Villalonga, B., & Amit, R. (2006). How do family ownership, control and management affect firm value? Journal of financial Economics, 80(2), 385-417.
Wang, X., Jiang, W., & Luo, Z. (2016). Combination of convolutional and recurrent neural network for sentiment analysis of short texts. Paper presented at the Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers.
Wang, Y., Huang, M., & Zhao, L. (2016). Attention-based LSTM for aspect-level sentiment classification. Paper presented at the Proceedings of the 2016 conference on empirical methods in natural language processing.
Zhang, Y., & Wallace, B. (2015). A sensitivity analysis of (and practitioners'' guide to) convolutional neural networks for sentence classification. arXiv preprint arXiv:1510.03820.