Querying on spatio-temporal databases and graphs

Abstract

With the proliferation of online social media (e.g., Facebook, Twitter, and Weibo), a huge amount of data with spatial, textual, and temporal dimensions is being generated at an unprecedented scale. Such spatio-textual data contains valuable information, and often reflects information dissemination. Users can query such data to retrieve their desired information. However, managing the data of large volume and efficiently processing the queries of various types bring great challenges to current database systems. To bridge the gap, we conduct the first study, which is to build a distributed system on streaming spatio-textual data (SSTD). It integrates multiple types of queries, and is equipped with a novel indexing structure for efficiency and load balancing strategies for robustness. In addition, to meet users’ various needs when retrieving the information of interest, we develop a novel type of query termed example-based spatial pattern matching (EPM) in our second study. Users can provide a set of spatial objects to serve as an example pattern, and search for the sets of objects from the database that exhibit a similar pattern as in the example. We propose efficient algorithms for solving EPM queries. Such data can also be represented as labelled graphs in some domain-specific applications. To enable interactive analysis in these applications, we propose time-aware attributes graph (TAG) summarization in our third study. TAG is such a graph that each vertex contains temporal, spatial and other optional attributes. Users pose queries on a TAG, and we return the summaries of the subgraphs satisfying the queries, which are beneficial for users to visualize and explore their interested data.