Desarrollo de un sistema de transcripción y contextualización automática para la generación de guías de estudio

Higher education faces persistent challenges in ensuring student accessibility and comprehension of content in theory-heavy courses. This project details the development of a web application designed to automatically transcribe class audio and contextualize the information to generate comprehensive study guides, thereby enhancing student learning across various demanding disciplines. The system leverages a Vosk (Kaldi) Automatic Speech Recognition (ASR) model, fine-tuned for Latin American Spanish and academic discourse, to produce accurate transcriptions. These transcriptions are subsequently enriched by integrating student-taken notes and open-access bibliographic resources. The core output is the automatic generation of structured, referenced study guides, exportable in Markdown format. Key technologies employed include Python, ChromaDB for vectorial data management, and JavaScript for the web interface. This initiative aims to provide an open-source, adaptable solution to improve understanding and academic performance in subjects with high conceptual density.

• Autores: Benítez Avilez, Felipe José; Gómez Rosales, Laura Sofía; Valencia Gómez, Fernando Mateo
• Fecha de publicación: 2025
• Institución: Universidad del Norte
• Repositorio: Repositorio Uninorte
• Idioma: spa
• OAI Identifier: oai:manglar.uninorte.edu.co:10584/13381
• Acceso en línea: http://hdl.handle.net/10584/13381
• Palabra clave: Automatic Speech Recognition (ASR), Natural Language Processing (NLP), Educational Technology, Higher Education, Contextualization, Fine-tuning, Study Guides, Latin American Spanish, Vector Search, Markdow; Reconocimiento automático del habla (ASR), Procesamiento del lenguaje natural (PLN), Tecnología educativa, Educación superior, Contextualización, Ajuste, Guías de estudio, Español latinoamericano, Búsqueda vectorial, Markdow
• License: Universidad del Norte