,As part of the research work, algorithms were developed to extract business rules from the content of document texts and an evaluator of these rules was developed to verify whether other related documents are correct in the context of the identified rules. The problem is related to the interpretation of so-called deontic expressions. These expressions have a structure that expresses in natural language certain rules that may or may not be fulfilled when certain events occur. Rule evaluation boils down to the calculation of a logical true/false value depending on the values of the variables associated with the event, which are evaluated according to the definition of the rule algorithm. In scientific work on the subject, a certain classification of these rules is usually adopted, depending on how their implementation affects the events occurring in a given context. Rule modality in the most general terms divides rules into permissive rules (allowing certain events), obligatory rules (specifying that certain events must occur) and invariant rules, which must always be true. Due to the distinctive area of operation, it was assumed that the domain context to be considered would primarily concern the validation of rules set out in contracts defining obligations to counterparties in relation to invoices issued between counterparties. An additional assumption was that the solution should primarily work for the Polish language, which is important for the market on which the Beneficiary operates. The problem thus defined was solved using natural language processing (NLP) techniques.,