Application of NLP Research Methods to the Analysis of Business Processes for Advertising Campaigns Implementation

The goal of this project is to create a data-driven approach to analyzing the business processes of advertising campaigns implementation, considering the specifics of the sphere, such as keeping all the processes in text form, which makes it impossible to analyze them using tabular methods. At the moment the communication of managers is tracked and analyzed manually, which is an inefficient way in terms of spending time. However, it is possible to solve these problems using natural language processing methods. This work answers the questions of how to evaluate the communication profile in a department using data, identify connections between employees, determine the number of mistakes and measure toxicity in communication and highlight features for further processing using tabular methods. For this project we collected textual data (comments from tasks in Bitrix24), pre-processed, applied clustering, statistical analysis, sentiment analysis, toxicity analysis, done thematic modeling, trained a model for counting mistakes in work, applied graph analytics and identified practices for feature extraction. Keywords: NLP, Bitrix24, machine learning methods, advertising campaign implementation, business process analytics.