“自然语言-草图”耦合的地理场景查询方法研究

Both natural language and virtual geographical scene are two important representation means of geographical scene. However, the integrality of geographical scene is ignored on the traditional spatial query, in which spatial objects, spatial relations and attributes are treated as query objects separately. This proposal aims to explore the method for geographical scene with the perspectives of geographical and geography. First of all, a classification schema of geography scene elements and a semantic relationship expression method are discussed based on the semantic web; Secondly, a geographical scene query frame is proposed to represent natural language query forms based on the frame knowledge representation method, and a visual method is present to create sketch maps of geographical scene inductively, in order to generate query representation models by means of natural language and maps. Thirdly, a similarity model is illustrated for measurement of geographical scene query matching, and a query ranking algorithm is developed. Finally, a prototype of geographical scene query system is developed for experimental evaluation. The research results could hopefully initiate a new method of geographic information query, deeply understand the mechanisms of cognition, representation and application of geographical scene, and extend the intelligence, popularization, and socialization of geographical information science and technology.

自然语言和图形是人们对地理场景认知和表达的两种重要手段，但是以空间对象、空间关系和属性为对象的空间查询模式忽略了地理场景的整体特性。本项目拟从地理认知和地理学的视角，研究地理场景组成要素的概念分类体系和语义关系表达方法；研究基于框架表示法的地理场景查询模式表达方法、引导式的可视化地理场景查询草图构建方法，构建基于自然语言和草图的地理场景查询表达模型；研究基于粗糙集理论的地理场景查询匹配相似度计算模型，设计自然语言和草图相结合的查询结果排序算法；研发地理场景查询原型系统，以矢量地理数据为实验数据源进行实验验证分析。研究成果有望形成一种地理信息查询和服务的新手段，进一步揭示地理场景的认知、描述和应用机理，促进地理信息科学和技术的智能化、大众化、社会化发展。

自然语言和图形是人们对地理场景认知和表达的两种重要手段，在表达的直观性、认知的习惯性、语义的丰富性等方面各有所长，互为补充。本研究以自然语言和草图两种地理场景表达手段为切入点，开展了自然语言和草图耦合的地理场景查询方法研究，突破传统以空间对象、空间关系和属性为对象的空间查询模式，建立一种新型的地理场景查询机制。研究开展过程中，围绕“概念表达—查询模式—计算模型”的技术主线，突破了地理场景概念分类体系、交互草图查询模式、地理场景驱动的相似性计算模型等关键技术，形成以下研究成果：（1）地理场景具有多层级嵌套、多要素相互作用的特性，在归纳分析了不同类型的地理场景的自然语言描述实例的基础上，采用语义网络对其进行概念建模，解决了地理场景结构化与非结构化描述共存、定性与定量结合等问题。（2）以草图绘制和信息标记为输入，提出了交互式场景的查询模式，并利用自然语言处理技术以及草图要素空间关系计算实现了查询场景的语义解析； （3）通过地理概念相似度、空间关系相似度和空间结构相似度计算，实现了基于多层次地理场景的相似计算方法，将地理场景查询转化为基于语义的场景相似性计算，从而实现地理场景的“文”、“图”到“景”的查询。本研究提供了一种地理信息空间查询的新手段，有益于揭示地理场景的认知、描述和应用机理。