This is a course project of the first semester of the master's program. The project is dedicated to the analysis of the question answering BERT-based systems. The project consists of two stages:

• The first (small) stage is devoted to studying the work of the BERT model with texts with different punctuation pre-processing.
• The second (bigger) stage contains the research of the model performance during work with texts of different lengths and texts with additional destructive information.