近年來，文字分析(textual analysis)的技術越來越成熟，主題模型(topic model)為其中一種文字分析方式，用於萃取文本的潛在主題(latent topic)。本研究使用潛在狄利克雷分布(latent Dirichlet allocation, LDA)主題模型及其延伸的情感主題混合模型(joint sentiment-topic model, JST)與反向情感主題混合模型(reverse joint sentiment-topic model, Reverse-JST)從10-K報吿文本中生成主題變數，結合財務比率變數，以羅吉斯迴歸模型(logistic regression model)方式，建構破產預測模型。
根據實證結果顯示，納入主題變數的破產預測模型能夠有效提升模型分類績效，且結合情感分析之主題變數更能助於優化預測模型，因而可以從 10-K 報告中的用詞觀察到是否企業破產的跡象。

In recent years, the technique of textual analysis has been well-developed. Topic modeling is part of a class of textual analysis methods, which extracts latent topics from documents. This paper uses LDA topic modeling and its extensions, JST and Reverse-JST, to generate topic-related variables from 10-K filings, and constructs corporate default prediction model in the form of logistic regression with topic-related variables and financial variables as independent variables.
According to the empirical results, when topic-related variables are included in the prediction model, the performance of classification is enhanced. In addition, considering sentiment analysis, topic-related variables are useful to optimize the prediction model. Therefore, by looking at the word usage of 10-K filings, investors can be aware of the sign of corporate bankruptcy.

第一章 緒論 1
第一節 研究動機與背景 1
第二節 研究目的 2
第二章 文獻探討 5
第一節 破產預測研究 5
第二節 主題模型 6
第三章 研究方法 9
第一節 主題模型 9
第二節 破產預測模型 21
第三節 模型績效衡量 22
第四章 資料來源與處理 26
第一節 10-K報告 26
第二節 情感詞典 26
第三節 財務變數 27
第五章 實證分析 29
第一節 破產預測模型建構 29
第二節 模型績效評估 42
第六章 結論與建議 58
參考文獻 60

Altman, E. I. (1968). Financial ratios, discriminant analysis and prediction of corporate bankruptcy. The Journal of Finance, 23(4), 589-609.
Black, F., & Scholes, M. (1973). The pricing of options and corporate liabilities, The Journal of Political Economy, 81(3),637-654.
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet allocation. Journal of Machine Learning Research, 3(Jan), 993-1022.
De Finetti, B. (1990). Theory of probability. Vol. 1-2. Chichester: John Wiley & Sons Ltd.
Deerwester, S., Dumais, S., Landauer, T., Furnas, G., & Harshman, R. (1990). Indexing by latent semantic analysis. Journal of the American Society of Information Science, 41(6), 391-407.
Duan, J. C., Sun, J. and Wang, T. (2012). Multiperiod corporate default prediction- A forward intensity approach, Journal of Econometrics, 170, 191-209.
Heinrich, G. (2005). Parameter estimation for text analysis. Web: http://www.arbylon.net/publications/text-est/pdf.
Hofmann, T. (1999). Probabilistic latent semantic analysis. Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence, Morgan Kaufmann Publishers Inc., 289-296.
Lin, C., He, Y., Everson, R., & Rüger, S. (2012). Weakly supervised joint sentiment-topic detection from text. IEEE Transactions on Knowledge and Data Engineering, 24(6), 1134-1145.
Lopatta, K., Gloger, M. A., & Jaeschke, R. (2017). Can language predict bankruptcy? The explanatory power of tone in 10-K filings. Accounting Perspectives, 16(4), 315-343.
Loughran, T., & McDonald, B. (2016). Textual analysis in accounting and finance: a survey, Journal of Accounting Research, 54(4),1187-1230.
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.
Merton, R. (1974). On the pricing of corporate debt: The risk structure of interest rates. The Journal of Finance, 29, 449-470.
Minka, T., & Lafferty, J. (2002). Expectation-propagation for the generative aspect model. Proceedings of the Eighteenth Conference on Uncertainty in Artificial Intelligence, 352-359.
Nguyen, T. H., & Shirai, K. (2015). Topic modeling based sentiment analysis on social media for stock market. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, 1, 1354-1364.
Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, 18, 109-131.