TitleThree essays on using text analytic techniques for accounting research
NameChakraborty, Vasundhara (author), Vasarhelyi, Miklos A (chair), Kogan, Alexander (internal member), Alles, Michael (internal member), Fisher, Ingrid (outside member), Rutgers University, Graduate School - Newark,
Accounting—Data processing ,
Disclosure in accounting
DescriptionThe wide variety of corporate as well as academic documents accessible online provides tremendous opportunities for data collection but at the same time presents a challenge for users to extract appropriate information from this immensely diverse corpus. Financial decisions are often based on different types of reports provided by firms including the financial statements as well as disclosures. To ensure simplified access to critical information from these reports data tagging is recommended, which creates the need for the development of taxonomies. A hybrid method of taxonomy creation using historical data from pension footnotes of 10K statements is proposed here instead of the pre-existing normative approach (detailed in the first essay). A comparison of the hybrid vs. normative approach (XBRL) pension footnote taxonomy reveals that generally firms tend to aggregate more while reporting, conflicting with findings from the XBRL pension taxonomy. Two experiments are conducted within the fold of the second essay to recognize whether (i) the type of industry (ii) firm size are factors significantly influencing the pension disclosure reporting of firms. It is found that while firm size is a significant factor the type of industry is not. Corporate reports as well as academic accounting literature are experiencing an unprecedented proliferation. The third essay explores the potential of using text analytic techniques to develop a methodology for automatic classification of academic articles in accounting based on different criteria. Three different experiments were conducted to seek the most effective method to accurately and automatically classify accounting literature. Results from the experiments indicate that automatic classification is more effective when using abstracts rather than using only keywords. Proposing a new hybrid method for taxonomy creation as well as the initial study that adopts automatic classification method in categorizing accounting literature by predetermined taxonomy classes is a contribution to accounting information systems literature. Understanding the impact of industry type and firm size on pension disclosure is a contribution towards accounting literature. Future research could be conducted with increased size of data corpuses, by experimenting with more advanced methods to replace all the manual steps with complete automation.
NoteIncludes bibliographical references
Noteby Vasundhara Chakraborty
CollectionGraduate School - Newark Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.