A Novel Framework for the Semiautomatic Generation of Research Protocols by Applying Transformer Network

This work focuses on the web framework development to semi-automatically generate the scope, limitations, and hypotheses of a research project through natural language processing. Initially, we paid special attention to the selection of the natural language processing architecture, conducting a comprehensive review of the state-of-the-art that covered methodologies, software platforms, and implementations in the field of natural language processing. Once the architecture was selected, we proceeded with designing the API responsible for processing requests from the web framework. Subsequently, the web framework was developed, where input data is captured, such as the proposed title, the issue, justification, and overall objectives. This data is obtained through the graphical interface of the proposed framework. The captured information is processed by a natural language processing server implemented in the Python programming language, using the GPT-2 model from OpenAI, based on the Transformer architecture. As a result, the manuscript of the research project proposal is generated in PDF format, demonstrating the feasibility of generating a research project through natural language processing.