The T2Know project explores the application of natural language processing technologies to build a semantic platform for scientific documents using knowledge graphs. These graphs will interconnect meaningful sections from different documents, enabling both trend analysis and the generation of informed recommendations. The project’s objectives include the development of entity recognition systems, the definition of user and document profiles, and the linking of documents through transformer-based technologies. Consequently, the extracted relevant content will go beyond standard metadata such as titles and author affiliations, extending also to other key sections of scientific articles, including references, which are treated as integral components of the knowledge representation.