Researchers of ancient cities often rely on textual descriptions, limited archaeological excavations, and scattered map materials to infer urban forms due to the absence of intuitive materials. However, urban descriptions in historical texts are often composed of interconnected information. The intricate and complex relationships between these pieces of information form a three-dimensional and multidimensional spatiotemporal model of the ancient city. In the conventional research context, researchers need to manually combine these historical materials to make reasonable conjectures, but such conjectures are often limited by the researcher’s personal data and view, making it difficult to form a rapid and precise generation mechanism. This research introduces big data processing and large language model technologies to construct an effective symbolic database framework for conveniently analyzing, demonstrating, and utilizing Chinese ancient city information in historical texts. This framework allows for the rapid generation of different “representations” of ancient city at specific historical moments, facilitating scholarly assessments of historical scenarios, which promotes the integration of computational technology, historical urban studies, and architecture, and offering new tools and perspectives for related academic work.