Natural language based posting account classification

Any process automation is meaningful when minimizing or excluding human involvement compared to the manual process. Operational accuracy plays a key role in accounting, so improving the accuracy of the results of the automation process will increase the willingness of accountants to use these solutions in their daily lives. This paper examines methods for the automation of the assignment of account codes to posting entries using natural language processing-based classification methods. We find that using textual comment features in the machine learning process allows to achieve debit account classification precision that is by up to 2.56 percentage points higher compared to previous work, reaching an overall classification precision of 94.78%.