Processing of Unstructured Data with GPT

This paper explores the use of artificial intelligence technologies for processing unstructured data from PDF files of scientific articles to automate key information extraction. Given the increasing volume of scientific literature, the development of effective techniques for analyzing and arranging the data contained within these articles has become highly important. The proposed approach uses PyMyPDF library to convert PDF content into a text format. The extracted text is then processed by a GPT-based neural network via an API, which analyzes and identifies key information such as the article title, authors' names, publication date, affiliated institutions, and other relevant details. The processed results are returned in JSON format, which simplifies further data analysis. The paper concludes with the development of a web service enabling automated extraction and analysis of data from scientific articles. This work has practical significance for researchers seeking to optimize the process of searching and analyzing scientific information, as well as for developers working in the field of natural language processing. The results of this work are implemented in the form of educational materials (lectures and practical classes) in Tomsk Polytechnic University in the course “Management of the digital ecosystem of the university” in the direction of “Management in scientific and educational sphere” for undergraduates.