摘要
为了降低与主题不相关的地名对地理相关性产生的负面影响,提出一种基于文档地名感知的排序方法。该方法不确定文档的地理范围,而是分别计算文档中出现的每个地名与查询范围的相关性,以减小文档中不相关地名对检索结果的影响。实验表明,基于文档地名感知的方法的检索效果优于确定最小边界矩形的方法和基于tf-idf的方法。
In order to induce the negative impact of irrelevant place names on geographical relevance, the authors proposed a geographic information retrieval (GIR) ranking method to reduce the impact which was based on document placenames-aware. Proposed method was not fixed on a MBR(minimum bounding rectangle), but all of the placenames which appeared in a document would calculate the relevance with query scope respectively. Experimental results show that the proposed approach can outperform baselines which are based on determining the MBR's model and pure textual model of tf-idf.
出处
《北京大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2013年第2期219-226,共8页
Acta Scientiarum Naturalium Universitatis Pekinensis
基金
国家自然科学基金(60673039
60973068)
国家社科基金(08BTQ025)
教育部留学回国人员科研启动基金
高等学校博士学科点专项科研基金(20090041110002
20110041110034)资助
关键词
地理信息检索
地理搜索
相关性排序
gographic information retrieval (GIR)
Geo-search
relevance ranking