Smart Chatbot Design to Support PDF Document Analysis at Politeknik Indonusa Surakarta

Abstract

The increasing volume of academic documents in PDF format presents challenges for students, lecturers, and academic staff in quickly accessing specific information. This study proposes the design and development of an intelligent chatbot system that facilitates semantic analysis of academic PDF documents at Politeknik Indonusa Surakarta. The system integrates Natural Language Processing (NLP) techniques and a Large Language Model (LLM), specifically GPT-4, using the Langchain framework to interpret user queries and deliver context-aware responses. The Research and Development (R&D) methodology was applied using a 4D model: Define, Design, Develop, and Disseminate. A prototype was developed with capabilities such as extracting content, summarizing sections, and answering user queries based on uploaded academic PDFs. Functional and usability testing were conducted using real academic documents. The results indicate high response accuracy (90%) and strong user satisfaction (score: 4.5/5), validating the system’s performance. The chatbot demonstrated its ability to support academic services by improving access to unstructured knowledge and streamlining information retrieval. Despite its potential, the system also faces challenges including PDF structure variations, dependency on third-party APIs, and the need for data privacy safeguards. This research provides a foundation for future implementations of AI-powered educational tools, suggesting further development such as multilingual support, voice interaction, and institutional integration.