DefinePK

Abstract

This research presents the development and evaluation of the "Offline Memory Mate Assistant," a secure AI-driven chatbot designed for personal data management. The main goal of this work is to develop a chatbot a secure AI-driven chatbot designed for personal data management, leveraging advanced natural language processing (NLP) specifically transfer learning and encryption techniques to ensure data privacy and protection. Utilizing the Stanford Question Answering Dataset (SQuAD2.0) for training, the research employs the pre-trained deep-set/roberta-base-squad2 model for fine-tuning contextual question-answering capabilities. The encryption method applied is Ferret from Python’s cryptography library, providing robust data security through end-to-end encryption. Additionally, the system is implemented with an offline mode to enhance privacy by processing all data locally, eliminating the need for internet transmission. A web-based user interface that is coded in HTML, CSS and JavaScript, Flask REST API for the model’s deployment and a Python script for data extraction and encryption. The interaction of the users with the chatbot is quite easy since it incorporates a user interface where the user can select document type and then receives the correct and meaningful response to the queries made. By addressing critical issues such as data privacy, encryption, and regulatory compliance, the study contributes valuable insights into the development of AI-driven systems that enhance user trust and satisfaction while ensuring data security.