Developing an AI-driven network suite : Leveraging LLMs for automated document processing and generation

Abstract

The development of AI-driven document generation systems has the potential to transform traditional manual report creation and management across various industries. Existing tools based on Large Language Models (LLMs) are often limited to text-based inputs and lack the ability to generate structured documents based on predefined templates, making them unsuitable for applications requiring the integration of multimodal data such as text, audio, and images. This limitation presents a significant challenge in domains like research and industry, where comprehensive, multimedia-rich structured documentation is essential. This study addresses these gaps by developing a user-friendly AI-driven document generation system that seamlessly integrates state-of-the-art LLMs, including OpenAI GPT-4o, Whisper, and Llama within a Learning Management Network Suite (LMNS). The system generates documents based on predefined templates, enabling consistency and efficiency. Custom evaluation metrics, ROUGE, BERTScore, and METEOR assess the generated document quality while load testing confirms the system's scalability. The study demonstrates that integrating multiple LLMs within a single system allows for the utilization of each model's strengths, ensuring optimal performance across various tasks, including text generation, summarization, audio transcription, and image analysis. Moreover, the incorporation of local editing and download functionality sets this work apart from existing solutions, offering users enhanced flexibility and control over the generated documents.