Business analytics using machine learning and large-scale textual data: Three essays

Natural Language Processing (NLP) techniques have been recognized as useful tools in various applications, such as language translation, stock price prediction, and semantic analysis. In recent years, the larger amount of data and faster computational hardware become available, NLP techniques are applied in a much broader application spectrum, and business applications are the frontier. My dissertation works at the intersection of business, Information Systems, and NLP. It extends NLP application scenarios in the business setting and employs NLP techniques in different real-world scenarios, including document topic categorization, company performance prediction, and corporate event sequence prediction. In particular, these problems are addressed by topic modeling techniques, recurrent neural networks, and sequence-to-sequence neural network architecture. My dissertation comprises three independent essays. The first essay utilizes topic modeling techniques to demonstrate the scientific topic trends and topic morphing for top Management Information System (MIS) publications. The second essay proposes a recurrent neural network model to predict multiple company financial ratios from publicly available news articles. The third essay first formulates a unique business question, predicting corporate event sequences, based on the historical event sequences submitted in the U.S. Securities and Exchange Commission (SEC) filings. Then, it uses the Transformer architecture to address the problem. In these essays, I apply and extend a variety of NLP tools to solve real-world business problems, including analyzing discipline publication trends and morphing, predicting company financial performance, and predicting corporate event sequences. NLP techniques employed in my dissertation include topic modeling, recurrent neural networks, and sequence-to-sequence networks. These essays contribute to discipline study and financial technology literature and develop tools for scholars, investors, and organizations to make informed decisions, from natural language