DescriptionThe objective of this dissertation is to investigate whether the sentiment features of business communication documents or social media information extracted by deep learning techniques deliver relevant and reliable information to auditors. The first essay investigates the incremental informativeness of sentiment features of earnings conference calls for the prediction of internal control material weaknesses (ICMW). With the help of a deep learning textual analyzer provided by IBM Watson, Alchemy Language API, this essay obtains the overall sentiment score of the text and the confidence score of the emotion “joy.” These sentiment features are then used as additional predictors along with other determinants of ICMW suggested by prior literature (i.e., Doyle, Ge, and McVay, 2007a; Ashbaugh-Skaife, Collins, and Kinney, 2007). The results indicate that these sentiment features, especially the score of joy, improve the explanatory ability and the prediction accuracy of the model. The second essay compares deep learning to the “bag of words” approach and demonstrates the effectiveness and efficiency of deep learning-based sentiment analysis for MD&A sections of 10-K filings in the context of financial misstatement prediction. The findings include (1) sentiment features provide insights for financial misstatement prediction, primarily for fraud detection; (2) the model using deep learning-based sentiment features generally performs more effectively than the model using sentiment features extracted by the “bag of words” approach. The third essay examines how the information of tweeting activities about the client company is associated with the audit fee. It examines the relationship between the audit fee of U.S. public firms in 2015 and the properties of tweets about the client firm: the sentiment of tweets, the volume of tweets, and the popularity of tweets. All tweet information is obtained using IBM Twitter Insights, a Twitter data analysis tool that provides sentiment and other enrichments relying on deep learning algorithms. It finds that for companies without going-concern audit opinions and companies with a median level of restatement risk, the audit fee is positively associated with the frequency of negative tweets, and this association is strengthened for companies receiving more retweets than those receiving less retweets.