基于空间场景相似性的投诉地址推荐

发布时间：2018-06-09 15:19

本文选题：类型本体 + 空间关系　；参考：《武汉大学》2017年硕士论文

【摘要】：随着地理信息技术的发展,海量的空间数据井喷式爆发,传统的数据检索模式已经不能满足人们对于空间数据的查询需求。例如要定位某个地址,人们往往是输入地址名称来查询,如果名称输入错误或者与空间数据库中的名称不匹配,通常得不到理想的结果。格式塔心理学指出,人们对于空间的认知是由整体到局部的,基于人们对空间的认知过程和描述习惯,本文提出了基于空间场景相似性的地址匹配模型。在用户不知道待查询地址精确名称的情况下,通过对待查询地址的空间关系进行描述,就可以检索到符合空间关系的地址,再进一步的选择自己要查询的地址。基于空间相似关系的空间数据检索,更贴近人们描述空间关系时的思维模式,对于未来空间数据智能化检索具有重要的意义。本文以12315投诉地址快速推荐为目标,研究了空间场景相似性的特征和计算方法。空间场景特征包括了两个方面:空间目标特征和空间关系特征。在研究空间目标特征时,本文提出了基于类型本体的语义相似度与基于编辑距离的字面相似度相结合的方法,检索用户输入的地名信息,以解决精确匹配得不到查询结果的问题。空间关系的相似性计算,主要涉及了空间方位、拓扑和距离关系。本文总结了三种空间关系的已有相似性计算方法,提出了本文所采用的空间关系计算模型。最后,基于空间目标特征和空间关系的相似性,提出了考虑局部重要性的空间场景相似性计算模型,并在原型系统中对该模型进行了可行性验证。论文的主要研究工作和成果如下:(1)提出基于类型本体的语义相似度与基于编辑距离的字面相似度相结合的地名匹配模型。针对12315投诉热线面临的地名地址无法精确匹配的问题,本文构建了类型本体,结合了字面相似度和类型语义相似度,提出了地名地址匹配模型。(2)提出空间关系相似性计算方法。本文研究的空间关系包括方位、拓扑和距离关系,在已有的空间关系相似性计算方法的基础上,针对12315投诉热线的应用情景和数据特征,提出了本文的空间关系相似性计算模型。(3)基于空间目标特征与空间关系,提出了考虑局部重要性的空间场景相似度计算模型。本文考虑了空间目标特征与空间关系两个方面,提出了空间场景相似性的计算模型。之后又考虑到不同空间目标对空间场景相似性影响的程度可能会不相同,本文对空间场景相似性计算模型进行了改进,允许用户定义空间目标对空间相似性的影响等级。(4)原型系统开发。通过对1235投诉地址快速定位的需求进行分析,本文提出了基于空间场景相似性的地址推荐的解决方案,并开发出原型系统,对本文提出的场景相似性计算模型的可行性进行验证。
[Abstract]:With the development of geographic information technology and the explosion of massive spatial data, the traditional data retrieval mode can not meet the demand of spatial data query. For example, to locate an address, people often enter an address name to query, and if the name is incorrectly typed or does not match the name in the spatial database, the desired result is usually not obtained. Gestalt psychology points out that people's cognition of space is from whole to part. Based on the cognitive process and description habits of space, this paper proposes an address matching model based on spatial scene similarity. If the user does not know the exact name of the address to be queried, by describing the spatial relation of the query address, the address that conforms to the spatial relationship can be retrieved, and the address to be queried is further selected. Spatial data retrieval based on spatial similarity relation is more close to the thinking mode of describing spatial relationship, which is of great significance to intelligent retrieval of spatial data in the future. In this paper, the feature and calculation method of spatial scene similarity are studied with the aim of fast recommendation of 12315 complaint address. Spatial scene feature includes two aspects: spatial target feature and spatial relation feature. When studying spatial target features, this paper proposes a method of combining semantic similarity based on type ontology and literal similarity based on editing distance to retrieve user input toponymic information. In order to solve the problem of accurate matching can not get the results of the query. The similarity calculation of spatial relation mainly involves spatial azimuth, topology and distance relation. In this paper, we summarize the existing similarity calculation methods of three kinds of spatial relations, and propose the model of spatial relation calculation used in this paper. Finally, based on the similarity of spatial target features and spatial relations, a spatial scene similarity calculation model considering local importance is proposed, and the feasibility of the model is verified in the prototype system. The main work and results of this paper are as follows: (1) A toponymic matching model based on semantic similarity based on type ontology and literal similarity based on editing distance is proposed. In order to solve the problem that the location and address of the 12315 complaint hotline can not be accurately matched, this paper constructs the type ontology, combines the literal similarity and the semantic similarity of the type, and puts forward a method to calculate the similarity of the spatial relationship by using the toponymic address matching model. The spatial relationships studied in this paper include azimuth, topology and distance relations. Based on the existing methods for calculating the similarity of spatial relationships, the application scenarios and data features of the 12315 complaint hotline are discussed. Based on the spatial object features and spatial relations, a spatial scene similarity calculation model considering local importance is proposed in this paper. In this paper, two aspects of spatial target feature and spatial relationship are considered, and a spatial scene similarity calculation model is proposed. After that, considering that different spatial targets may have different effects on spatial scene similarity, this paper improves the spatial scene similarity calculation model. Allows the user to define the impact of spatial objects on spatial similarity level. 4) prototype system development. By analyzing the requirement of fast location of 1235 complaint address, this paper puts forward a solution of address recommendation based on spatial scene similarity, and develops a prototype system. The feasibility of the scene similarity calculation model proposed in this paper is verified.
【学位授予单位】：武汉大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F203;P208

【相似文献】