A hybrid method for taxonomy creation

Abstract

Automatic extraction of information from the footnotes of financial statements can be particularly challenging due to a wide variation in filing structure and terminologies. Standardized text and use of data tagging can facilitate this process. This paper: (i) proposes and demonstrates a new hybrid method of taxonomy creation, using historical data; (ii) compares the taxonomy structure using the new method, with that of the existing XBRL US GAAP taxonomy; (iii) shows evidence of structural differences between the official XBRL US GAAP taxonomy and the new hybrid taxonomy and (iv) demonstrates how the tool so developed could be used for more exploratory research. Comparison of this new structure with that of the existing XBRL taxonomy structure reveals that its creation from historical data provides a greater level of aggregation compared to the XBRL US GAAP taxonomy for Pension footnotes.