Linked data publication pipelines for agri-related use cases

This document is part of a project that has received funding
from the European Union’s Horizon 2020 research and innovation programme
under agreement No 732064. It is the property of the DataBio consortium and shall not be distributed or
reproduced without the formal approval of the DataBio Management Committee. Find us at www.databio.eu.
1
This project has received funding from
the European Union’s Horizon 2020
research and innovation programme
under grant agreement No 732064
This project is part
of BDV PPP
LINKED DATA PUBLICATION PIPELINES
FOR AGRI-RELATED USE CASES
Raul Palma1, Soumya Brahma1 , Marcin Krystek1, Karel
Charvát2, Raitis Berzins2
1Poznan Supercomputing and Networking Center, Poland
2WirelessInfo, Czech Republic

7th Leipzig Semantic Web Day
22nd May, 2019

2
Linked data publication
•  LD is increasingly becoming a popular method for publishing data on the Web
•  Improves data accessibility by both humans and machines, e.g., for finding, reuse and integration
•  Enables to discover more useful data through the links (and inferencing), and to exploit data with semantic
queries
•  Growing number of datasets in the LOD cloud
•  1,234 datasets with 16,136 links (as of June 2018)
•  Coverage of the LOD cloud
•  Large cross-domain datasets (dbpedia, freebase, etc.)
•  Variable domain coverage (e.g., Geography,
Government, BioInformatics)
•  What about Agriculture?
•  “Just” few datasets (e.g., AGRIS biblio records,
AGROVOC thesaurus + other thesaurus like NALT)
•  Farming and agri-activities related data?
http://lod-cloud.net/

3
Why is Linked Data relevant in Agriculture:
Farming context
•  Farm management
•  Multiple activities and stakeholders
•  Multiple applications, tools and devices
•  Multiple data sources, types and formats
• Challenge
•  To combine/integrate those different and
heterogeneous data sources in order to
make economically and environmentally
sound decisions

4
Data Integration in relevant projects (context)
•  Data integration challenges have been the focus of relevant projects
EU FP7, ICT CIP, 2014- 2017
EU FP7, ICT CIP, 2014- 2017
FOODIE aimed at building an open
and interoperable cloud-based
platform addressing among others the
integration of data relevant to
farming production including their
geo-spatial dimension, as well as their
publication as Linked data.

SDI4Apps aimed at building a cloud-
based framework with open API for
data integration focusing on the
development of six pilot apps, drawing
along the lines of INSPIRE, Copernicus
and GEOSS
DataBio aims at showcasing the benefits of Big
Data technologies in the raw material production
from agriculture & others for the bioeconomy
industry; deploying an interoperable platform on
top of the existing partners’ infrastructure.
DataBio aims at delivering solutions for big data
mgmt., including i) the storage and querying of
various big data sources; ii)
the harmonization and integration of a large
variety of data from many sources, using linked
data as a federated layer

5
Linked data principles & guidelines
•  Simple set of principles & technologies
•  URI, HTTP, RDF, SPARQL
•  Involves a set of (common) tasks
Datasets
identification
Model specification
RDF data generation
Linking
Exploiting
Hyland et al.
Hausenblas et al.
Villazón-Terrazas et al.
Best Practices for Publishing Linked Data
5-star deployment scheme
for Linked Open Data

6
Implementing Linked Data publication pipelines

7
Implementing Linked Data publication pipelines
•  Goal: to define and deploy (semi-) automatic processes to carry out the necessary steps to transform and
publish different input datasets as Linked Data.
•  A pipeline connect different data processing components to carry out the transformation of data into RDF
and their linking, and includes the mapping specifications to process the input datasets.
•  Each pipeline is configured to support specific input dataset types (same format, model and delivery
form).
•  Principles
•  Pipelines can be directly re-executed and re-applied
(e.g., extended/updated datasets)
•  Pipelines must be easily reusable
•  Pipelines must be easily adapted for new input datasets
•  Pipeline execution should be as automatic as possible.
The final target is to fully automated processes.
•  Pipelines should support both: (mostly) static data and data streams
(e.g., sensor data)
•  The resulting datasets available as Linked Data, will provide an integrated view over the initial
(disconnected and heterogeneous) datasets, in compliance with any privacy and access control needs

8
Use cases (examples)
•  UC1: Publication of farm related linked data from agri pilots in DataBio
project, particularly Pilot 8 [B1.4] Cereals and biomass crops.
•  Goal: query and access heterogeneous agricultural data sources from Rostenice farm via an integrated
layer in order to make informed decisions and discover new knowledge
•  UC2: Publication of Open EU/national datasets relevant for agri-food pilots
as Linked Data
•  Goal: provide access to multiple, isolated data sources relevant for agri-pilots, and identify links with
farm datasets, from a single integrated layer
•  UC3: Publication of sensor data as linked data on the fly from Pilot 9 [B2.1]
Machinery management
•  Goal: provide access to sensor data integrated with other farm data and other relevant datasets

9
Datasets identification and collection
•  UC1 Datasets
•  Farm data (Rostenice holding) that holds information
about each field names with the associated cereal
crop classifications and arranged by year.
•  Data about the field boundaries and crop map and
yield potential of most of the fields in Rostenice farm
from Czech Republic.
•  Yield records from two fields (Pivovarka, Predni)
within the pilot farm that were harvested in
2017/2018.
Source data types: shapefiles

10
•  UC2 datasets
•  Czech datasets
•  Czech LPIS data showing the actual land boundaries.
•  Czech erosion zones (strongly/SEO and moderately / MEO erosion-endangered soil zones)
•  Czech water bodies (e.g., restricted area near to water bodies has 25m buffer according to
the nitrate directive).
•  The data about soil types from all over Czech Republic.
•  Polish datasets
•  Polish LPIS data showing the cadastral land boundaries from all over the country.
•  European datasets
•  FADN (Farm Accountancy Data Network) data about the income of agricultural holdings and
the impacts of the Common Agricultural Policy from all EU member states
•  Various open European geospatial datasets including
•  (part of) Open Land Use (OLU)
•  (part of) Open Transport Map (OTM)
•  Smart Points of Interest (SPOI),
•  (part of) Urban Atlas (pan-European comparable land use and land cover data for Large
Urban Zones )
•  (part of) CORINE Land Cover
•  HILUCS (Hierarchical INSPIRE Land Use Classification System )
•  Experimental sample dataset from the review platform Yelp (global coverage).
The data contents are regarding the geographical location of a business, review
and reviewer information.
Source data types: shapefiles, JSON, CSV,
relational databases

11
•  UC3 dataset
•  Sensor data stored and collected by Senslog platform,
including the readings of IoT devices on tractors.
•  Data updates is high, reading coming frequently
Source data types:
relational databases

12
Data models
•  Various ontologies/vocabularies were selected and reused/extended for the
representation of data
•  For farm-related data, INSPIRE based FOODIE ontology has been selected and extended as needed
•  For the land parcel and cadastral data (for Czech republic & Poland), erosion-endangered soil zones, water buffer
and soil type classification, also FOODIE ontology and in some cases its extensions were used for modelling the
classes and properties.
•  In case of the FADN the main ontologies used were Data Cube Vocabulary and its SDMX ISO standard extensions
that were much effective in aligning such multidimensional data. Moreover the Data Cube Vocabulary
encompasses well known RDF vocabularies like SKOS, SCOVO, VoiD, FOAF, Dublin Core, etc.
•  For the Yelp dataset various ontologies like review, FOAF, schema.org, POI, etc. were used to represent the
classes and the properties identified from the input data sources.
•  For some other datasets (e.g., corine, hilucs, olu, otm, urban atlas) simple ontologies/voabularies were
generated in line with standards and are available in https://github.com/FOODIE-cloud/ontology.
•  For sensor data: Semantic Sensor Network (SSN) along with the SOSA (Sensor, Observation, Sample, and
Actuator) ontology for describing sensors and their observations, the involved procedures, the studied features of
interest, etc. Additionally, Data Cube Vocabulary and its SDMX ISO standard extensions for multidimensional data

13
Data model for farming data
•  (FOODIE) farming data model principles
•  application vocabulary covering the different categories of information
dealt by the farm mgmt. tools/apps (in FOODIE)
•  in line with existing standards and best practices
•  Resulting model*
•  Builds on the INSPIRE AF specification for agricultural data, and
•  the INSPIRE specification for themes in annex I for geospatial data, based on
•  ISO/OGC standards for geographical information
•  Created as an UML model
*consulted with experts from various institutions, e.g., EU DG JRC, EU Global
Navigation Satellite Systems Agency (GSA), Czech Ministry of Agriculture,
Global Earth Observation System of Systems (GEOSS), German Kuratorium für
Technik und Bauwesen in der Landwirtschaft (KTBL).
Challenge: Transform model into OWL
ontology

14
Transformation of model into
OWL ontology
Palma R., Reznik T., Esbri M., Charvat K., Mazurek C., An INSPIRE-
based vocabulary for the publication of Agricultural Linked Data.
Proceedings of the OWLED Workshop: collocated with the
ISWC-2015, Bethlehem PA, USA, October 11-15, 2015
ShapeChange implements ISO 19150-2
standard rules for mapping ISO geographic
information UML models to OWL
ontologies.
semi-automatic process: besides
transformation configuration,
additional pre and post processing
task were needed

15
FOODIE ontology - overview
•  Overall structure (ShapeChange output)
•  UML featureTypes and dataTypes modelled as classes, and
their attributes as datatype or object properties
•  UML codeLists modelled as classes/concepts, and their
attributes as concept members
•  Cardinalities restrictions defined on properties (exactly,
min, max)
•  DataType properties ranges defined according to model/
mappings
•  Object properties ranges defined according to model/
mappings
•  Object properties inverseOf defined
Top hierarchy
FeatureType hierarchy
Codelist hierarchy
Datatype hierarchy

16
FOODIE ontology – main classes overview
•  We found the lack of a feature on a more detailed level than Site that is already part of
the INSPIRE AF data model.
•  Main concept: Plot
•  Represents a continuous area of agricultural
land with one type of crop species, cultivated
by one user in one farming mode
•  Two kinds of data associated:
•  metadata information
•  agro-related information
§  Next level: Management Zone
•  Enables a more precise description of the land
characteristics in fine-grained area
foodie:Plot
INSPIRE-AF:Site
foodie:Alert Foodie:Intervention Foodie:CropSpecies
Foodie:ManagementZone
containsPlot
containsManagementZone
interventionPlot speciesPlot
alertPlot
plotAlert
Foodie:ProductionType
production
Foodie:SoilNutrients
zoneNutrients

17
FOODIE ontology – main classes overview
•  The Intervention is the basic feature type for any kind of (farming)
application with explicitly defined geometry, e.g., tillage or pruning.
•  Has multiple indirect associations with different concepts
Foodie:Intervention
Foodie:Treatment
Foodie:TreatmentPlan
Foodie:Product Foodie:ProductPreparation
Foodie:ActiveIngredients
is-a
plan
productPlan
planProduct
preparationProduct preparation
productTreatment
treatmentProduct
preparationPlan
ingredientProduct
Foodie:FormOfT
reatmentValue
Foodie:Treatme
ntPurposeValue
formOfTreatment treatmentPurpose

18
RDF data generation
•  Different tools were deployed, configured and used for the generation of RDF
data, depending on the source data type
•  D2RQ: mainly for relational databases
•  Geotriples: mainly for shapefiles
•  R2MLprocessor: mainly for JSON, CSV data sources
•  All these tools require a mapping file (in RDF) specifying how to map the data
source elements to the target ontology concepts and properties.
•  Mapping specifications use R2RML (RDB to RDF Mapping Language) and/or
extensions (e.g., RDF Mapping language (RML))

19
RDF data generation
•  The mapping file also specifies the connection details for the source dataset

•  Based on the mapping file, the data source (e.g., database content, shapefile, JSON,
CSV, etc.) is either
•  i) dumped to an RDF file; or
•  ii) transformed on the fly as a virtual RDF graph (e.g., for data streaming)
•  RDF files were loaded into Virtuoso triplestore (FOODIE endpoint)

20
Linking the generated RDF datasets
•  In order to link the resulting RDF datasets with
other datasets, we follow different approaches:
•  Apply existing tools like Silk or LIMES to discover
equivalence relations
•  For other relations use queries (e.g., geospatial) to
access the integrated data as per need.
•  In our experiments with equivalence relations;
however we also had to do some manual
entries
•  We found issues in handling large datasets in Silk,
specially those accessed via SPARQL endpoint that
we cannot control
There were recent optimizations of LIMES
and a new tool for geospatial linking (from
Leipzig) that we plan to test

21
Triplestore statistics (examples)
Dataset Name Graph in FOODIE endpoint Source Triples
OLU** http://w3id.org/foodie/olu# Transformed from PostgreSQL 127,926,060
SPOI http://www.sdi4apps.eu/poi.rdf Provided by WRLS (also available
in FOODIE endpoint)
407,629,170
NUTS http://nuts.geovocab.org/ Open Source (available in
FOODIE endpoint)
316,238
OTM*** http://w3id.org/foodie/otm# Transformed from PostgreSQL 154,340,785
Yelp academic
dataset
http://data.yelp.com/academic_dataset# Transformed from JSON 86,348,185
LPIS data (CZ) http://w3id.org/foodie/open/cz/
pLPIS_180616_WGS#
Transformed from shapefile 24,491,282
FADN http://ec.europa.eu/agriculture/FADN/
{dataset}#
Transformed from CSV 25,457,255
Pilot 8 farm data private Transformed from shapefile 1,569,439
Total: over 1 bilion triples!
FOODIE triplestore is one of the largest semantic repositories related to agriculture, which has been
recognized by the EC innovation radar as „arable farming data integrator for Smart Farming”

22
Exploiting the Linked Data – querying triplestore
•  Sparql endpoint: https://www.foodie-cloud.org/sparql

23
Exploiting the Linked Data – querying sensor data as
virtual RDF graph
•  Sparql endpoint: http://senslogrdf.foodie-cloud.org/sparql
•  SNORQL search endpoint: http://senslogrdf.foodie-cloud.org/snorql/
•  Web-based visualization: http://senslogrdf.foodie-cloud.org/

24
Exploiting the Linked Data – querying examples(1-2
linked datasets)
•  get information of Points Of Interest in a
given polygon (one dataset)
•  get SPOIs for given Open Land Use parcel
(linking 2 datasets)
SELECT * FROM <http://www.sdi4apps.eu/poi.rdf> WHERE { ?Resource rdfs:label ?Label . ?Resource poi:class ?POI_Class . ?Resource geo:asWKT ?Coordinates . FILTER(bif:st_intersects (?Coordinates,
bif:st_geomFromText("POLYGON((6.11553983198 54.438016608357,
6.95050076948 47.230985358357, 13.36651639448 47.626493170857,
14.99249295698 54.701688483357, 6.11553983198
54.438016608357))"))) . }
SELECT * FROM <http://www.sdi4apps.eu/poi.rdf> WHERE { ?Resource rdfs:label ?Label . ?Resource poi:class ?POI_Class . ?Resource geo:asWKT ?Coordinates . FILTER(bif:st_intersects (?Coordinates, bif:st_geomFromText(?
coordinates))) . { SELECT bif:st_astext(?x) as ?coordinates FROM <http://w3id.org/foodie/olu#> WHERE { olu-instance: geo:hasGeometry ?geometry. ?geometry geo:asWKT ?x } }
}

25
Exploiting the Linked Data – querying examples (3+
linked datasets)
•  Show me all the land parcels (OLU) that
have hotels (SPOI) and that lie not more
than 50 meters away from the major
highway (OTM) (linking 3 datasets)?
SELECT DISTINCT ?olu ?hilucs ?source ?municode ?specificLandUse FROM <http://w3id.org/foodie/olu#> WHERE { ?olu a olu:LandUse . ?olu geo:hasGeometry ?geometry . ?olu olu:hilucsLandUse ?hilucs . ?olu olu:geometrySource ?source . OPTIONAL {?olu olu:municipalCode ?municode} . OPTIONAL {?olu olu:specificLandUse ?specificLandUse} . ?geometry geo:asWKT ?coordinatesOLU . FILTER(bif:st_within(bif:st_geomFromText(?coordinatesPOI),?coordinatesOLU)). { SELECT DISTINCT ?Resource, ?Label, bif:st_astext(?coordinatesPOIa) as ?coordinatesPOI FROM <http://www.sdi4apps.eu/poi.rdf> WHERE { ?Resource rdfs:label ?Label . ?Resource poi:class <http://gis.zcu.cz/SPOI/Ontology#lodging> . ?Resource geo:asWKT ?coordinatesPOIa . FILTER(bif:st_within(?coordinatesPOIa,bif:st_geomFromText(?coordinatesOTM),0.00045)) . { SELECT bif:st_astext(?x) as ?coordinatesOTM FROM <http://w3id.org/foodie/otm#> WHERE { ?roadlink a otm:RoadLink . ?roadlink otm:roadName ?name. ?roadlink otm:functionalRoadClass ?class. ?roadlink otm:centerLineGeometry ?geometry . ?geometry geo:asWKT ?x . FILTER(bif:st_intersects (?x, bif:st_geomFromText("POLYGON((14.426647 50.0751251,14.426647 50.07685089,14.4305469
50.07685089,14.43054696 50.0751251,14.426647 50.0751251))"))) . FILTER(STRSTARTS(STR(?class),"firstClass") ) . } } } }
}

26
Exploiting the Linked Data – search & navigate
•  Faceted search endpoint: http://www.foodie-cloud.org/fct

27
Exploiting the Linked Data – visualisation
•  Map visualisation: http://ng.hslayers.org/examples/foodie-zones/

28
•  Map visualisation: http://ng.hslayers.org/examples/produce-3d/
Object information
on click

29
•  Map visualisation: http://ng.hslayers.org/examples/olu_spoi
•  OLU polygons colored by the number of SPOI that lie inside them

30
•  Map visualisation: http://app.hslayers.org/project-databio/land
•  Usage scenarios:
•  Find and show buffer zones around water bodies (user will specify the distance),
which define the areas within the fields with limited/restricted application of agro-
chemicals.
•  select farm/fields based on the ID_UZ attribute from public CZ LPIS database and
search EO data over all fields
•  visualize crop species based on the farm data (need parcels with crop types- not
available from open LPIS data)
•  select fields with different soil types
•  select all fields with certain crop in max distance from certain point (it could be for
logistic, distribution of biomass etc)
•  show/select erosion zones for specific farm ID

31
•  Map visualisation: http://app.hslayers.org/project-databio/land

32
•  Metaphactory: http://foodie.metaphacts.cloud/resource/Start

33

34

35

36
Security of RDF Data
•  SPARQL endpoints are web services capable of providing Read-Only access to a back-end graph DBMS.
•  SPARQL endpoints can be sometimes purpose-specific so their access privilege therefore must be limited
to some basic operations over the graph.
•  The privileges provided by a given Virtuoso SPARQL endpoint include specific user identities with specific
database roles and privileges.
•  Virtuoso offers three methods for securing SPARQL endpoints:
•  Digest Authentication via SQL Accounts
•  OAuth Protocol based Authentication
•  WebID Protocol based Authentication

37
Linked data publication technologies overview
•  Used technologies:
•  D2RQ, Geotriples and R2MLProcessor for input
datasets as Virtual RDF Graphs
•  RDF for the representation of data
•  FOODIE (Farming ontology) providing the underlying
vocabulary and relations, plus a number of other
ontologies/vocabularies (existing and generated)
•  Virtuoso for storing the semantic data
•  Silk/LIMES for discovery of links
•  Sparql for querying semantic data
•  Hslayers NG for visualisation of data
•  Metaphactory for visualisation of data
D2RQ

38
Thank you for your attention!
Raul Palma <rpalma@man.poznan.pl>

Linked data publication pipelines for agri-related use cases

Recomendados

Recomendados

Más contenido relacionado

La actualidad más candente

La actualidad más candente (12)

Similar a Linked data publication pipelines for agri-related use cases

Similar a Linked data publication pipelines for agri-related use cases (20)

Más de Leipziger Semantic Web Tag

Más de Leipziger Semantic Web Tag (17)

Último

Último (20)

Linked data publication pipelines for agri-related use cases