Κατεύθυνση Υπολογιστικά Συστήματα: Λογισμικό και Υλικό
Πληροφορική

2021-11-24

2021

Papamichalopoulos Marios

Μανόλης Κουμπαράκης, Καθηγητής, Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ

JedAI-spatial: a system for 3-dimensional Geospatial Interlinking

English
Greek

JedAI-spatial: a system for 3-dimensional Geospatial Interlinking

Geospatial data constitutes a considerable part of Semantic Web data, but so far, its sources lack enough links in the Linked Open Data cloud. Geospatial Interlinking aims to cover this gap by associating geometries with established topological relations, such as those of the Dimensionally Extended 9-Intersection Model. Various algorithms have already been proposed in the literature for this task.

In the context of this master thesis, we develop JedAI-spatial, a novel, open-source system that organizes the main existing algorithms according to three dimensions:

i. Space Tiling distinguishes interlinking algorithms into grid-, tree- and partition-based, according to their approach for reducing the search space and, thus, the computational cost of this inherently quadratic task. The former category includes Semantic Web techniques that define a static or dynamic EquiGrid and verify pairs of geometries whose minimum bounding rectangles intersect at least one common cell. Tree-based algorithms encompass established main-memory spatial join techniques from the database community, while the partition-based category includes variations of the cornerstone of computational geometry, i.e., the plane sweep algorithm.

ii. Budget-awareness distinguishes interlinking algorithms into budget-agnostic and budget-aware ones. The former constitute batch techniques that produce results only after completing their processing over the entire input data, while the latter operate in a pay-as-you-go manner that produces results progressively - their goal is to verify related geometries before the non-related ones.

iii. Execution mode distinguishes interlinking algorithms into serial ones, which are carried out using a single CPU-core, and parallel ones, which leverage massive parallelization on top of Apache Spark.
Extensive experimental evaluations were performed along these 3 dimensions, with the experimental outcomes providing interesting insights about the relative performance of the considered algorithms.

Technology - Computer science

Geospatial Interlinking, Massive Parallelization, Filter-Verification Framework

Yes

5

Yes

3

76

https://pergamos.lib.uoa.gr/uoa/dl/object/2967186/file.pdf

https://pergamos.lib.uoa.gr/uoa/dl/object/2967186