一种优化的自然语言空间查询转换模型  

An optimized model for translating natural language to GeoSQL

作　　者：亢孟军[1,2] 曹浩杰 苏世亮 翁敏[1] 王明军[1] KANG Mengjun;CAO Haojie;SU Shiliang;WENG Min;WANG Mingjun(School of Resource and Environmental Sciences,Wuhan University,Wuhan 430079,China;Beijing Key Laboratory of Urban Spatial Information Engineering,Beijing 100038,China)

机构地区：[1]武汉大学资源与环境科学学院,武汉430079 [2]城市空间信息工程北京市重点实验室,北京100038

出　　处：《测绘科学》2022年第7期194-200,共7页Science of Surveying and Mapping

基　　金：城市空间信息工程北京市重点实验室经费资助项目(20210211)

摘　　要：针对将自然语言描述的空间查询问题转换为可执行的SQL语句的研究热点,提出了一种优化的自然语言空间查询转换模型,采用BERT预训练模型以及空间命名实体识别技术标注空间查询问题中的地理实体以及空间算子(包括取坐标、求距离、缓冲区),将生成的词向量输入SQL输出模块,构建空间查询SQL语句。使用英文语料生成SQL语句并进行了精度检验,发现模型的准确率较高,并且相比于基线模型准确性有着明显的提升。实验结果表明:所提模型能够利用自然语言问题快速准确地查询空间数据,研究结果可以帮助普通用户在不提前学习专业空间数据库知识之前的前提下提高人机交互效率,促进地理信息向智能服务方向发展。According to the research front of translating spatial queries in natural language into computer-executable SQL queries,in this paper,a new natural language to GeoSQL translating model is delivered.First,this model annotates geographic entities and spatial operators through the word embedding model of BERT and the spatial named entity recognition model.Second,generated word vectors are sent to the SQL output module to build GeoSQL.The SQL statements are generated in the English corpus.The accuracy inspection indicates that this translation model did a good performance in generating spatial SQL statements,and this model has exceeded the performance of other baseline models.This model makes improvements for generating spatial SQL queries faster and accurately through natural language queries.Users with this model can use spatial databases for better human-computer interaction without knowing specialized spatial database knowledge in advance.This model aims at supporting geographic information services into information services.

关 键 词：自然语言查询接口 BERT 命名实体识别 空间算子 地理实体 

分 类 号：P208[天文地球—地图制图学与地理信息工程]