(The) effects of lexical features on information quality by data analytics

Abstract

Individuals gather information online to resolve problems in their daily lives and make various decisions about the purchase of products or services. With the development of the Internet, web 2.0 has allowed more people to easily generate and use online reviews such that the volume of information is rapidly increasing, and the usefulness and significance in analyzing the unstructured data have also increased. Recently, methodologies that can be used to analyze such unstructured data have been introduced because of the development of analytical technology. Text mining is one of these new methodologies. This thesis uses text mining to analyze the types of information that individuals can access, and determines how this information is used when individuals make decisions. In Essay 1, I focus on narrative sections of annual reports, which must be mandatorily disclosed in financial reporting, to reveal the relationship between company assessment information written directly by managers and the performance of companies using the text mining approach. To analyze this relationship, I collected the annual reports of all the publicly listed firms in the United States from 1996 to 2011 and conducted text mining to identify the tones of the annual reports and observe whether the tones of the annual reports changed depending on the current earnings level. In addition, I explore factors that could cause tone flexibility in the reports. I compare companies whose tones in their reports were more positive with those with less positive tones compared to their current performance to analyze how the future performance of these two groups would be different compared to the current performance. In Essay 2, diverse lexical features of expert reviews and user review contents provided by a third-party review site are extracted and defined. Specifically, the lexical properties of the product reviews are defined according to each relevant factor. Then, the influence of each linguistic factor of expert reviews and user reviews on consumers’ final evaluations is tested. In addition, product price is applied as a contextual variable to classify products into high-price products group and low-price products group, and test whether each linguistic factor influencing consumers’ evaluations towards products is different depending on lexical features. Finally, I investigate whether the lexical features of reviews and quantitative evaluations, rated by experts and users, affect consumers’ purchase decisions. Through these analyses, I expect to provide guidelines for how individuals process massive volumes of unstructured data depending on lexical features in various contexts, and how companies can use this mechanism from their perspective.