This master’s thesis explores the integration of Knowledge Graphs (KGs) with Large Language Models (LLMs) to enhance the interpretability and usability of Building Information Modeling (BIM) API documentation. The research addresses challenges in understanding complex and unstructured documentation by transforming it into a queryable format using KGs, tested with the Vectorworks API and 36 user queries. A hybrid approach combining deterministic graph construction and LLM-generated embeddings proved effective, ensuring reliable relationships, adaptability to unstructured data, and improved accuracy in text and code suggestions. The study demonstrates how graph-based Retrieval-Augmented Generation (RAG) agents leverage the semantic richness of KGs to answer queries efficiently, bridging the gap between developers and advanced BIM tools. This approach improves productivity and decision-making in construction workflows. While limitations like data quality and computational demands of embeddings are noted, the research highlights the potential of hybrid KG construction for enhancing API documentation interpretation. It advocates further exploration of KGs and LLMs to advance BIM workflows in the construction industry.     «

This master’s thesis explores the integration of Knowledge Graphs (KGs) with Large Language Models (LLMs) to enhance the interpretability and usability of Building Information Modeling (BIM) API documentation. The research addresses challenges in understanding complex and unstructured documentation by transforming it into a queryable format using KGs, tested with the Vectorworks API and 36 user queries. A hybrid approach combining deterministic graph construction and LLM-generated embeddings pro...     »