Leveraging Geming Pro And Bert For Automated Imrad Classification: A Novel Dataset And Saas Platform

Résumé: This thesis presents a novel microservices-based SaaS platform for the automated anal-ysis of scientific paper introductions. The platform utilizes advanced natural language processing (NLP) techniques to accurately cassify sentences within introductions ac-cording to their rhetorical purpose, following the widely recognized IMRaD (Introduc-tion, Methods, Results, and Discussion) structure. The research involved a multi-stage process of developing and refining AI models,culminating in a custom-built dataset and highly accurate BERT models fine-tuned for specific IMRaD classification tasks. The platform's robust microservices architec-ture,leveraging technologies such as Next.js, Spring Boot, Python,TensorFlow Serving,Express.js, Redis, and Nginx, ensures scalability, maintainability, and optimal perfor-mance. This platform provides users with an intuitive interface for uploading research pa-pers, analyzing introductions, visualizing results, and offering valuable feedback. Pre-mium features, including automated summarization and a hypothetical reconstruction of the author's thought process, are available through a flexible subscription model. This thesis demonstrates the feasibility and effectiveness of using AI to analyze IMRaD structure, offering a valuable too for students, researchers, and educators to improve their understanding and practice of scientific writing. The platform's modular design and open-source codebase lay a strong foundation for future development and expansion in the field of scientific communication.

Publié dans la revue: