Utilizing Open-Source AI to Navigate and Interpret Technical Documents : leveraging RAG models for enhanced analysis and solutions in product documentation

Abstract

This thesis investigates the application of Retrieval-Augmented Generation (RAG) systems for the analysis of extensive technical product manuals, particularly in the context of VEO, a company specializing in electrification and automation solutions. RAG systems represent an advanced integration of natural language processing technologies, combining the strengths of dense retrievers and sophisticated embedding models. This synergy facilitates the effective extraction of relevant information from large-scale document collections, followed by the generation of accurate, context-aware responses.

Focusing on the challenges faced by VEO, this thesis outlines how RAG models are tailored to streamline the retrieval and analysis of multiple iterations of product manuals from various vendors. The study details the operational framework of RAG models and demonstrates their effectiveness in enhancing the accessibility and interpretability of technical documentation. Through a systematic exploration of the model architecture and practical applications, this research highlights the transformative potential of RAG models in improving problem-solving processes within technical fields, thereby boosting productivity, and reducing operational risks associated with manual errors. The study can also help clarify the technical requirements, installation processes, and potential challenges involved in integrating AI solutions into VEO's infrastructure.