Efficient Time-interval Aware Spatial Keyword Queries On Road Networks

With the rapid development of location-based services and geo-positioning technologies,more and more spatial-textual objects which encompass a geo-location and a textual description are available.Spatial keyword queries become more and more popular in the fields of spatial-temporal database.Some technologies have been studied.However,most of these works only focus on textual and spatial information,but ignore the temporal information and only applicable to European space,while in many practical problems spatial-textual objects and queries are in road networks,and users usually need more information,such as the opening time and price to make a better decision.Existing work and related technologies cannot make full use of temporal information to solve user's query needs.This paper study the problem of efficient Time-interval Aware Spatial Keyword(TASK)queries which consider the location,time-interval and the attribute values of keywords of the spatial objects on road networks.The main contributions can be summarized into three points.First,this paper proposes the concept of keyword hot value which is usually the value of the textual attributes,such as popularity,price and rating.And designs a novel similarity function which considers spatial,textual and temporal dimension for TASK query to evaluate the similarity between spatial textual objects and queries.Second,a hierarchical index GI-tree is developed to organize the spatial textual objects efficiently.The distance matrix(DM and SC)keeps the position information,and the keyword information list(KAI)contains the textual information,the textual attribute values and temporal information in GI-tree.A similarity function is proposed to estimate the score between a GI-tree node and the query.Based on these,a baseline method(BM)with the best-first strategy is elaborated to process TASK queries.Experiments demonstrate its efficiency and indicate it possess applicable value.Third,to improve the query efficiency,a novel index structure termed SGI which is composed of SBT-trees and GI-tree is designed to prune unqualified objects by utilizing their spatial,textual and temporal information simultaneously.The spatial textual objects are effectively organized in corresponding SBT-trees based on their temporal information.Furthermore,based on these,an efficient prune strategy is elaborated to shrink the search space and a more reasonable similarity function is proposed to estimate the score between a GI-tree node and the query,as well as the heuristic searches are presented to get further optimization,and the search framework is designed to obtain the top-k results in an efficient way.Extensive experiments on real datasets are conducted to demonstrate the scalability and efficiency of our proposed solution based on SGI index.