The SPADES project is a research project that is funded by the Hellenic Foundation for Research and Innovation (HFRI) and the General Secretariat for Research and Technology (GSRT), under grant agreement No [1667]. The instrument aims to support post-doctoral research, and the principal investigator of SPADES is Akrivi Vlachou. The project is hosted at the Department of Digital Systems in the University of Piraeus.
SPADES won the Best Research Paper Award at SSTD'21 with the paper: A Novel Indexing Method for Spatial-Keyword Range Queries
Find Out MoreSeveral research challenges are associated with the efficient support of spatio-textual processing at scale. Two main factors are critical for location-based services, thus determining their overall performance: the efficiency of query processing for spatio-textual queries and the quality of the retrieved points of interest. the efficiency of query processing directly influences the query throughput, which is very important in the context of scalable applications.
Consequently, this project proposal identifies a set of research and technological challenges that need to be effectively addressed, in order to support spatio-textual and spatio-temporal queries over massive data:
The spatio-textual query returns objects that satisfy a spatial constraint and are also described by textual descriptions that best match the query keywords. The simplest form of spatial constraint is that the location of the retrieved object is in a given region on the current map, similar to a range query. Nevertheless, other spatial constraints, such as the nearest points of interest or those that are better than any other point of interest closer to the user location, do make sense for spatial-keyword search and they are not supported to-date. Even though a wide variety of spatial queries has been studied for spatial data, only a limited set of basic spatial constraints has been studied with respect to spatio-textual search. To alleviate this shortcoming, SPADES will define new advanced query types that allow the formulation of useful queries with complex constraints. It is important to support also queries that rank points of interest based on the textual descriptions that characterize other interesting facilities in their spatial neighborhood.
Efficient multi-dimensional access methods (such as R-trees) for spatial queries have been well researched in spatio-temporal databases in the past. Similarly, index structures for storage of textual information (documents) have been successfully developed by the IR-community in the past. The majority of such index structures have been developed independently. However, supporting efficient spatial-keyword search requires the combination of the merits of both worlds, and this has not been adequately studied yet for massive datasets that exceed the computational capacity of a single node. As such, distributed indexing is necessary, which is going to combine global with local indexes, suitable for spatiotextual data. This is also tightly related to issues such as data partitioning and load balancing, which also need to be investigated for spatio-textual data and under the respective query workloads.
In order to provide a generic and portable framework for parallel processing of spatio-textual data, SPADES will propose the design of a framework that consists of generic operators (e.g., filter, scan, index, distribute, etc.) that work on spatio-textual data. In this way, SPADES will put in place an abstraction for the definition of parallel processing algorithms, which can be customized for specific parallel processing engines, by providing the necessary implementation of the operators. Our intention is to decouple the algorithm specification from the underlying parallel processing engine, which is a methodology that is going to offer added value to our research. To demonstrate its feasibility, SPADES is going to provide implementations of all operators for a specific data-parallel processing engine.
The results of the project are expected to be exploited by the local tourism business that could benefit from the provision of innovative location-based services over vast quantities of data to tourists. Such services entail queries that are processing-intensive and typically run for minutes rather than seconds, and often produce results that are not truly useful to the end user. SPADES promises to facilitate the access to location-based information, a task of particular importance to tourists. In this respect, the research results of SPADES are expected to benefit also society at large.
SPADES aims to address the limitations of spatio-textual data analysis and processing when applied in the context of Big Spatial Data, as witnessed by the lack of existing systems and techniques for this purpose. Achievement of this goal constitutes a substantial step forward in dealing with challenges emerging from management of Big Spatial Data. At a practical level, the research outcome will benefit applications such as spatio-textual search and retrieval, mining of spatio-textual data, next generation location-based services, and tourism-oriented applications to name a few.
By exploiting SPADES the analysis of massive spatio-textual datasets (typically encountered in the aforementioned domains and especially in social networks) is going to be accelerated significantly. In consequence, applications will be able to query and analyze larger quantities of spatio-textual data in shorter time, thus speeding up the process of making new discoveries as well as aiding the task of interpretation of heterogeneous data (spatial or multidimensional data and unstructured textual data).
For more information please contact Akrivi Vlachou. For more information about the research group and the department, please visite the respective home pages: Department of Digital Systems
