An Accounting Classification System Using Constituency Analysis and Semantic Web Technologies

Abstract

To aid accountants in making professional judgments and decisions regarding the accounting methods for transactions, we propose a classification system by integrating computational linguistics with semantic web technologies. We use constituency parsing to convert the classification rules in accounting standards into a machine-processable data structure: Resource Description Framework (RDF) triples. When an accounting classification question is input, the system converts it into an RDF triple, compares it with the established triples of different accounting methods, and subsequently identifies the most appropriate accounting method. We showcased and evaluated our proposed model using IFRS 9 and IAS 28. Our study provides both scholarly and practical applications by (1) incorporating computational linguistics and semantic web technologies to create an interpretable, process-traceable, and explainable classification system aligned with regulatory requirements; and (2) proving that the knowledge-based model can be established without substantial training data, enhancing its accessibility and utility for accounting professionals.