Research Seminar: Mapping under LLM: Two Hierarchical Solutions

Cartography includes tasks of map making and map applications. Both have deep associations with Artificial Intelligence either in judgement decision or generation decision (AIGC). The Large Language Model, as a powerful deep learning technology, has driven the development of cartography in multiple impacts. Based on some practical works, this presentation will discuss the question LLM+Mapping from two hierarchical perspectives. One is using matured LLM tool, such as ChatGPT, builds extended mapping technology. The geo-text process supported by ChatGPT is inserted into mapping system to conduct mapping task understanding, text data extraction and annotation matching. Some intelligent map generation based on text automatic understanding has been realized by experiments. The second is developing some generation tool using map data and AIGC model, including by GAN to generate hill-shading map transferring cartographer’s experience, by GCN model to conduct map pattern recognition, and others. Two hierarchical solutions play different roles in settling question LLM+Mapping to drive cartography development.