Journal of Geo-information Science

Research and Implementation of Travel Itinerary Chain Extraction Method Based on the DeepSeek Model

CHEN Minjie, ZHANG Zheng, CAO Yibing, ZHANG Jiangshui, YANG Zhenkai, LU Zhenglun

Journal of Geo-information Science, 2026, 28(3): 722-735. DOI: 10.12082/dqxxkx.2026.250487

存在问题	示例
使用别名或简称	“万岁山武侠城”和“万岁山大宋武侠城” “明堂天堂遗址”和“天堂明堂” “洛阳古代艺术博物馆”和“古墓博物馆”
拼写错误	“山陕甘会馆”误写成“山陕甘会所” “明堂天堂”误写为“名堂天堂”
表述歧义	“真不同”实际上指“真不同博物馆”
整体与部分	“猕猴谷”、“茱萸峰”均属于“云台山景区” “白园”和“香山寺”均属于“龙门石窟景区”
同名异物	“人民公园”、“城隍庙”在不同城市都有

Tab. 2 Problems in the extraction results of travel node names

Other figure/table from this article

Fig. 1 Overall flow chart of travel itinerary chain extraction
Fig. 2 Travel itinerary chain description model
Fig. 3 Prompt strategy for generating travel itinerary chain data
Tab. 1 Detailed content and description of data format constraints
Fig. 4 Travel itinerary chaingeolocationprocess
Tab. 3 Examples of fields in an external knowledge file
Fig. 5 Temporal distribution of the extracted travel itinerary chains
Fig. 6 Spatial distribution of the extracted travel itinerary chainsin Henan Province
Fig. 7 Spatial distribution of travel nodes by different platforms in Henan Province
Fig. 8 Top 20 travel nodes by route frequency
Fig. 9 The number of 5A and 4A scenic spots in each city of Henan Province
Fig. 10 Scatter plot of the number of 5A and 4A-level scenic spots versus the frequency of travel nodes
Tab. 4 Macro average of precision, recall, and F1 score for extracted travel nodes （%）
Tab. 5 Similarity statistics of extracted travel itinerary chains with different lengths （%）
Tab. 6 Comparison of travel node extraction results （%）
Tab. 7 Extracted travel itinerary chain similarity comparison （%）