Chatbot basado en procesamiento de lenguaje natural para asistencia y recomendaciones en delitos cibernéticos en la Unidad Nacional de Ciberdelito

Abstract

Abstract: The design, development, and evaluation of the chatbot based on natural language processing for cybercrime assistance in the National Cybercrime Unit has been a successful process. It began with a comprehensive literature review on chatbots, identifying their evolution and applications in complex contexts, then began the design of the conversational protocol and a knowledge base tailored to the needs of citizens allowing the chatbot to provide clear answers on the prevention and reporting of cybercrime. For the development, advanced tools used such as OpenAI, which allowed the chatbot to understand and generate coherent responses in natural language, Python for API integration and logic management, Flask as a framework for the web application, Jinja2 to generate dynamic content in the responses and MySQL to store data efficiently. The system was evaluated through functional, performance, usability, and stress tests, obtaining positive results; with a score of SUS 73.3, it was validated that the chatbot has an acceptable usability, which implies that most users consider it easy to use, as well as proving that the system is efficient in terms of query management and its ability to operate 24/7. The final recommendations include improving chatbot customization, expanding the knowledge base and optimizing the scalability of the system, so overall it concluded that the project contributes to improving accessibility and effectiveness in the fight and prevention of cybercrime.