A Comparison of Information Architectures of Software Documentation in RAG-Based Delivery Scenarios

This paper examines the relevance of information architectures of content used for RAG-based systems within the domain of software documentation. It specifically focuses on how information architectures impact the retrieval performance of such documentation. For this purpose, the content is prepared using automated data processing, and its performance is evaluated through reuse metrics derived from previous research in technical communication. Furthermore, the paper investigates the effects of different truncation strategies and embedding models on retrieval performance. Finally, the sensitivity of the reuse metric MSD is evaluated by conducting a robustness test.