In this article, we propose the design, construction, and validation of a technological solution with the ability to automate the process of reviewing and analyzing thesis paper by using natural language processing and a deep learning training algorithm. This project seeks to be a proof of concept because current solutions in the academic field focus on abstracting and analyzing scientific articles but not in theses. Another point to consider is that these solutions are in English language. The resulting language model was compared with other language models based on the Transformers architecture. The result of this comparison gives us an objective for future research on the A.S.T.R.A. project.