LLM and RAG-Based Question Answering Assistant for Enterprise Knowledge Management

Large language models (LLM) have become integral to many natural language processing applications, particularly in the area of automatic question answering. In this study, a question answering system was developed to enable Adesso Turkiye employees to access internal company information quickly and accurately. A Retrieval Augmented Generation (RAG)-based question answering framework was constructed by utilizing multiple large language models and embedding techniques, along with content curated by experts in human resources and information security. The performance of the system was evaluated using ROUGE, BLEU, and accuracy metrics, and the results indicated high levels of success. Future work will focus on enhancing performance through the use of different language models, enriching the system with datasets from various domains, and integrating the developed system into MS Teams to ensure accessibility for employees.