Back to search

University of the Aegean

Infrastructure for discovering geographically related linked data

Abstract

dc:description

The provision of suitable tools to professionals, scientists and others, that would supportthe search for geographic data in the web and their integration, is a basic premise for thecompletion of many tasks such as map-making and spatial analyses. New ways of publishingsemantically-enriched data in the web according to the linked data principles result in therealization of a single global and interconnected web of data, which eases the accessibility,process and integration of huge amounts of data from various sources and offers advancedsearching and querying capabilities. In this research, we address the problem of discoveringrelated geographic data in the web of data. In particular, we propose a dataset recommendationfor link discovery methodology that recommends, to a given spatial dataset, other spatialdatasets in the web of data that may contain related entities, that is, entities that refer to semanticallyclose or to the same real-world geographic objects. The recommended datasetscan be then used as input to a link discovery process for the establishment of links (e.g.,sameAs) between their related entities. In contrast to the current approaches, which determinedataset relevancy based on the textual and structural similarity of datasets or by examiningexisting links, we approach the problem from a geographic perspective based on thehypothesis that “datasets whose entities present similar spatial distribution are likely to containsemantically related entities”. To support our hypothesis, we implemented a practical solutionthat operates in web scale and first creates summaries on geographical elements (pointentities) of spatial datasets provided through SPARQL endpoints and then apply metrics tocompute their similarity. The summaries capture geographical characteristics of datasets,such as their spatial extent and the spatial distribution of their entities. Various set-similarity,statistical and information theory metrics are applied on dataset summaries to compute a geographicalsimilarity score for datasets. A recommendation algorithm ranks relevant spatialdatasets according to their geographical similarity score, so as the top ranked datasets aremore probable to contain related entities with the queried spatial dataset. We conducted experimentsto evaluate the effectiveness and the efficiency of the dataset recommendation forlink discovery methodology. The results show that the proposed recommendation algorithmgenerate ranked lists with 62% mean average precision, approximately 35% higher than simplebaselines. Also, it reduces approximately 99% the search space for relevant data sourcesin the web of data using efficient techniques. A sideline contribution of our research concernsthe development of methods for the seamless access and parsing of web of data and thegeneration of reports that shed light on the size and the state of the geographic semantic webthat show that approximately 39% of the semantic web datasets contain georeferenced information.In addition, we propose GeoVoID, a metadata standard for describing geographiccharacteristics of datasets, such as their spatial extent, the spatial ontologies they use andthe number of their geographic entities. The outcome of our research is realized in a webapplication, called GeoLOD, an exhaustive catalog of spatial datasets in the semantic web and an online dataset recommender for link discovery. GeoLOD allows map-based searchfor datasets and offers novel features such as live preview of datasets geographic entities onan interactive map, export of configuration files for instant use in the Silk and LIMES linkdiscovery frameworks and on-the-fly recommendation for unknown SPARQL endpoints andspatial datasets in GIS compliant formats (e.g., Shapefile). The usefulness of GeoLOD forGIS professionals and linked data experts is confirmed by the results of a conducted userstudy.

Degree

thesis:*
Grantor dc:publisher
University of the Aegean
Year dc:date
2021

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Kopsacheilis, Vasileios
  • Κοψαχείλης, Βασίλειος

Subjects

dc:subject × 18

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/50780

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Kopsacheilis, Vasileios; Κοψαχείλης, Βασίλειος. Infrastructure for discovering geographically related linked data. University of the Aegean, 2021. http://hdl.handle.net/10442/hedi/50780