Using Semantic Similarity to Proxy for Knowledge Structures in Accounting Research
Journal
Journal of Information Systems
Journal Volume
39
Journal Issue
2
Start Page
41
End Page
70
ISSN
0888-7985
1558-7959
Date Issued
2025-05-20
Author(s)
Abstract
Accountants and auditors frequently make judgments that require complex evaluations of facts and rules. Researchers examine these judgments by conducting experiments in which participants make accounting decisions and then describe the information they used and how they used it in making their decisions. These elicited descriptions provide researchers insight into participants’ knowledge structures. This study uses a design science approach to develop a semantic similarity (a natural language processing analytic method) measure of how closely participants’ knowledge structures match an exemplar knowledge structure derived from authoritative accounting sources. The study then tests the semantic similarity metric in an experimental setting with two accounting tasks: assessing inventory obsolescence and production process business risk. Results show that the semantic similarity metric evaluates the knowledge structure match comparably to other methods requiring human judges. Importantly, the semantic similarity metric yields comparable results more efficiently.
Subjects
accounting judgement
accounting research
design science
knowledge structure
semantic text similarity
textual analysis
unstructured data
Publisher
American Accounting Association
Type
journal article