Research of Summarization Methods for Presentation Slides Generation

Making presentations is a creative process that requires the use of analytical skills to highlight the main thoughts and convey them in a concise and visually pleasing form. In this paper, popular language models were used for summarizing voluminous documents in Russian and compared on prepared datasets such as GusevGazeta, scientcific articles datasets in two languges: russian and english. With the help of SummaRuNNer extraction method, it was possible to achieve better accuracy of summarization of large russian texts' volumes than with other abstract methods under consideration. Proposed FastAPI application architecture for generating presentations. The service parses PDF files, extracts titles and images, creates a summary of the text using artificial intelligence.