Spatial Keyword Query Based On Negative Keywords And Semantic Processing

With the development of multimedia network and GPS global positioning service system and the popularization of related applications,a large number of multidimensional data has shown explosive growth,including a large number of spatial text object data with keyword attributes.At present,spatial database query processing technology is mainly used to analyze and process these data.The most important core issue is the spatial keyword query technology,which can quickly return to the user's points of interest to meet the query needs by utilizing the spatial and text attributes of the points of interest itself and considering their correlation with the query.Existing spatial keyword query technologies mainly focus on the distance calculation between latitude and longitude coordinates and accurate text matching,ignoring the keyword attributes excluded by users in the query,and only rely on the exact matching between the text in the query process.For queries in which the text information and the point of interest information cannot match exactly,it is very likely that no results will be returned within a certain area.This paper makes the following improvements:(1)To solve the problem of ignoring user exclusion attribute in query,this paper presents a new query mode with negative keyword constraint.First,Geohash strings are used to represent points of interest objects,B+-tree is inserted as the leaf nodes of the binary tree after sorting the strings,and the objects with negative keywords are filtered by the binary tree.A hybrid index structure BGIB-Tree based on Geohash is constructed.On the basis of this index and based on the recursion of Geohash encoding,a prefix matching search algorithm is designed,which uses the pruning strategy of prefix matching between area encoding and object encoding,so that objects satisfying spatial constraints can be quickly found without distance calculation through string matching,and the query can be completed by bidirectional search of the IDs of these objects in the inverted index.This can effectively handle the negative keyword information that users enter in the query.Compared with the R-tree method,the accuracy of this algorithm is improved by 29%.(2)For the phenomenon of polysemy or user input errors,it is easy to cause problems that can not return the most accurate query results to users.This paper presents a clustered hybrid index tree(GB2-Tree,Geohash Binary BC-i Distance Tree)that combines spatial,semantic and text hierarchies.First,the dimension reduction algorithm is used in the spatial dimension to optimize the pruning effect,and then BC-i Distance index is used in the semantic layer to cluster the high-dimensional semantic vectors,which can quickly and accurately find objects similar to the query semantics according to their subject distribution distance.The average query efficiency of this algorithm is 19.6% higher than that of NIQ algorithm.