Research On Reverse Collective Spatial Keyword Querying

The rapid spread of GPS devices and the rapid development of location-based services have produced a large number of spatial objects with textual information.As a variant of spatial keyword query,Collective Spatial Keyword Querying(CoSKQ)has become one of the research focuses currently.In this paper,we propose and define a new problem Reverse Collective Spatial Keyword Querying(RCoSKQ).In this kind of query,for any user in the result set,compared with other spatial object sets that also cover the query keyword set,the query point set has the highest spatial similarity with the user.However,in the query process of RCoSKQ,we face two problems.First,there are many spatial object sets covering the query keyword set,and an efficient method is needed to filter out the unqualified object set.Second,a fast method is required to filter qualified users.The main works of this paper as follows.In this paper,we first propose a regions-based pruning query algorithm.First,all the col-lective keyword sets are obtained.For each collective keyword set,we divide the space centered at each query point.Then,we prune the space of each partition and generate a partition influ-ence set.Finally,we use the partition influence set to filter qualified users.The intersection of qualified users of all query points is taken,and the qualified users of all the collective keyword sets are merged.Then,a half-space pruning based query algorithm is proposed based on the regions-based pruning query algorithm.First,all the collective keyword sets are obtained.For each collective keyword set,the collective influence region is calculated.Then,we use the collective influence region to filter the qualified users,and the users in the region are returned.Finally,we merge all the users of collective keyword sets.Finally,in this paper,we use two real data sets and several synthetic data sets to design reasonable comparison experiments to study the effects of different parameters on the response time and I/O cost.The experimental results show that the regions-based pruning query algorithm can effectively solve the RCoSKQ problem.But the half-space pruning based query algorithm is more efficient.