Semantic organization for historical maps: Classification, representation, association
针对历史地图语义混乱、丢失和歧义问题，提出包含分类、表示和关联机制的语义组织系统，并构建问答系统验证其有效性，在精确率和召回率上优于百度文心一言和GPT-4o。
Abstract Given that historical maps (HM) are represented by a complex network of symbols, their semantics cannot be easily and directly understood. To extract the embedded knowledge, scholars have developed semantic organization for different types of HM. However, the construction of semantic organization for HM is challenging due to problems of semantic clutter, semantic loss, and semantic ambiguity. To resolve these problems, this paper proposes a semantic organization system which includes classification, representation, and association mechanisms for HM. The intent is to achieve semantic ordering, semantic enhancement, and semantic association. As a means to verify the proposed semantic organization system, this paper develops an HM knowledge question and answer (Q&A) system. Experimental results show that the Q&A system outperformed Baidu (Wenxinyiyan) and GPT‐4o in terms of precision and recall.