基于知识图谱的自然语言式地名查询方法研究  

Research on natural language place name query method

作　　者：李诚 毛曦[1] 路文娟[1] 王继周[1] 董文洁 LI Cheng;MAO Xi;LUWenjuan;WANG Jizhou;DONG Wenjie(Chinese Academy of Surveying&Mapping,Beijing 100036,China)

机构地区：[1]中国测绘科学研究院,北京100036

出　　处：《测绘科学》2024年第4期187-195,共9页Science of Surveying and Mapping

基　　金：中央级公益性科研院所基本科研业务费项目(AR2308,AR2121)。

摘　　要：针对现有自然语言地名查询时存在的对查询意图理解不准确,查询过程中地名错查、漏查等歧义问题,该文提出了一种基于知识图谱的自然语言式地名查询算法。基于地名数据进行统计分类,并建立了相应的地名知识图谱,为查询提供了数据支撑;在分析常见的空间查询语言特点的基础上提出了一种有效的自然语言式空间查询的解析方案,提高了查询的智能性;针对地名简称、别名以及未录入词分别采用考虑字面相似度、位置信息、专通名影响的地名语义相似度算法,解决了地名查询中容易出现的一些错查漏查问题。实验结果表明,本文所提出的查询方法能理解多数情况下用户提出的查询语句并返回给用户较为准确的结果。To address the issues of inaccurate understanding of the query intent,incorrect and missed search of place names during the query process,a natural language based place name query algorithm was proposed based on knowledge graph in this paper.Based on place name data for statistical classification,a corresponding map of place name knowledge was established,providing data support for queries;On the basis of analyzing the characteristics of common spatial query languages,an effective parsing scheme for natural language based spatial queries was proposed,which improved the intelligence of queries;A composite place name semantic similarity algorithm was proposed for place name abbreviations,aliases,and non-entered words,which solved some of the problems of errors and omissions in place name queries.Experimental results showed that the proposed query method could understand the query statements proposed by users in most cases and return more accurate results to users.

关 键 词：知识图谱 自然语言处理 语义相似度 地名排歧 

分 类 号：P208[天文地球—地图制图学与地理信息工程]