One of the critical issues in the analysis and design of work systems is the updating of work method descriptions and work process controls. Changes in work methods may originate from product or process engineers or from employees involved in production. Work methods should be documented and incorporated into company documentation in accordance with quality assurance procedures. Therefore, it is necessary to develop a method that supports the creation of production documentation and employs an efficient data analysis technique, such as neural networks. This article presents the application of a deep neural network (DNN) to develop verbal descriptions of production processes based on video recordings. The conversion of visual data into verbal descriptions utilizes a predefined vocabulary of the assembly process. The structure of the DNN and the results of experiments are presented. The proposed approach is useful in developing best practices for identifying working methods tailored to specific needs.