Automating the Process of Taxonomy Creation and Comparison of Taxonomy Structures

18 Pages Posted: 5 Dec 2010

Date Written: October 10, 2010

Abstract

The ability to automatically extract information from the footnotes of financial statements simplifies access to critical information concerning public companies. However, extraction can be particularly challenging due to great variations in the filing structure and terminologies used. Hierarchical formalization of text becomes a necessity in such circumstances. This is facilitated by the creation of a valid taxonomy. The objectives of this paper are threefold: (1) to develop a semiautomatic method of taxonomy creation; (2) to compare the structure of the taxonomy created with the XBRL US GAAP taxonomy and (3) to demonstrate how the tool developed as a part of this process can be used for more exploratory research. Pension plan footnotes of 10K statements have been used to demonstrate the use of this method. To create a taxonomy, we first collected 10K statements from SEC EDGAR (Electronic Data Gathering, Analysis and Retrieval) then extracted pension footnotes, and restructured the data. We then applied the Hierarchical clustering algorithm to this data to create the taxonomy structure. Comparison of the taxonomy developed with the XBRL taxonomy reveals some differences. In general the reporting trends of companies reveal a greater level of aggregation. Pension footnote structures of forty five randomly selected companies were compared across ten years. Several instances were found where the company has added new terms or a completely new section to the footnote or a term is missing.

The contribution of this paper are: (i) a method to formalize and partially automate the complex and time consuming process of taxonomy creation using historical data has been proposed; (ii) a generic parsing tool as a part of the taxonomy creation process is developed; (iii) structural differences between the official XBRL US GAAP taxonomy and taxonomy using historical data are demonstrated (iv) potential use of the parsing and matching tool for other exploratory research in accounting is shown.

Keywords: Taxonomy, Pension disclosure, XBRL

JEL Classification: M4, M41

Suggested Citation

Chakraborty, Vasundhara and Vasarhelyi, Miklos A., Automating the Process of Taxonomy Creation and Comparison of Taxonomy Structures (October 10, 2010). Available at SSRN: https://ssrn.com/abstract=1719611 or http://dx.doi.org/10.2139/ssrn.1719611

Vasundhara Chakraborty (Contact Author)

Rutgers University ( email )

180 University Avenue
Newark, NJ 07102
United States

Miklos A. Vasarhelyi

Rutgers Business School ( email )

180 University Avenue
Ackerson Hall, Room 315
Newark, NJ 07102
United States
973-353-5002 (Phone)
973-353-1283 (Fax)

Register to save articles to
your library

Register

Paper statistics

Downloads
183
Abstract Views
891
rank
167,495
PlumX Metrics