=Paper=
{{Paper
|id=Vol-2262/ekaw-poster-11
|storemode=property
|title=A Use Case of Data Integration in Food Production
|pdfUrl=https://ceur-ws.org/Vol-2262/ekaw-poster-11.pdf
|volume=Vol-2262
|authors=Liliana Ibanescu,Thomas Allard,Stéphane Dervaux,Juliette Dibie,Elisabeth Guichard,Caroline Penicaud,Joe Raad
|dblpUrl=https://dblp.org/rec/conf/ekaw/IbanescuADDGPR18
}}
==A Use Case of Data Integration in Food Production==
https://ceur-ws.org/Vol-2262/ekaw-poster-11.pdf
A Use Case of Data Integration in Food
Production
Liliana Ibanescu1? , Thomas Allard2 , Stéphane Dervaux1 , Juliette Dibie1 ,
Elisabeth Guichard2 , Caroline Pénicaud3 , and Joe Raad1
1
UMR MIA-Paris, AgroParisTech, INRA, Université Paris-Saclay, 75005 Paris,
France
2
CSGA, AgroSupDijon, CNRS, INRA, Université Bourgogne Franche-Comté, 21000,
Dijon, France
3
UMR GMPA, AgroParisTech, INRA, Université Paris-Saclay, 78850,
Thiverval-Grignon, France
Abstract. This paper presents a use case about knowledge represen-
tation and integration of data from different domains in food science.
An ontology named PO2 DG, the Process and Observation Ontology for
the production of Dairy Gels, has been designed in order to provide a
shared vocabulary for domain experts. The available data have been se-
mantically structured using PO2 DG and are stored in an RDF repository
named PO2 DG dataset. This use case identifies some of the challenges
when dealing with a multi domain representation problem, gives some
hints about possible solutions and suggests some further work.
Keywords: process representation, experimental observations represen-
tation, food science, ontology based data integration
1 Introduction
One of the most significant current discussions in food production is how to
formulate food products with high nutritional and sensory values and low en-
vironmental impact. This supposes being able to build a decision support tool
combining data and knowledge from different domains in food science (e.g. nu-
trition, sensory and perception, eco-design, microbiology, biochemistry, process
engineering) with data and knowledge in environmental analysis.
The aim of the ongoing NutriSensAl project involving experts from INRA,
the French National Institute for Agricultural Research, is to propose a proof
of concept for a decision support tool allowing the formulation of well-balanced
products in terms of nutritional requirements (e.g. less fat, sugar and salt) with
acceptable sensory qualities for the consumer while using eco-friendly produc-
tion processes. While a lot of data have been collected in different collaborative
?
This work was supported by the Qualiment Carnot Institute of the French National
Research Agency through the NutriSensAl project (grant number 16CARN002601)
and by the Saclay Data Center through the LIONES project.
2 L. Ibanescu et al.
research projects, a major problem is the lack of cross domain studies fully com-
bining nutritional and sensory properties with eco-design processes.
The knowledge representation task of the NutriSensAl project focuses on the
production process of French hard cheeses, while studying consumer sensory per-
ception. In order to address data and knowledge integration, a relevant solution
is the use of an ontology. Therefore, PO2 DG, the Process and Observation Ontol-
ogy for the production of Dairy Gels, has been designed in order to provide a
shared vocabulary for the different domain experts involved in the project. The
ontology has been developed using Scenario 6 of the NeON methodology [1],
i.e. reusing, merging and re-engineering ontological resources. PO2 DG was imple-
mented in OWL 2, is available on AgroPortal and some of its concepts are aligned
with concepts from foundational ontologies or concepts from available ontologies
on the LOD (Linked Open Data) cloud [2]. An RDF repository PO2 DG dataset
stores the available data of the NutriSensAl project. The definition of a common
vocabulary and the use of Semantic Web technologies should facilitate access to
other similar data in order to reuse and integrate them in the future NutriSensAl
decision support tool.
The paper is organized as follows. In Section 2 the context of the NutriSensAl
project and the available data are presented. Section 3 gives details about the
data and knowledge representation task. In Section 4 we give some comments
about our experience during this project, and, finally, we conclude in Section 5
and present our further work.
2 NutriSensAl Project
The NutriSensAl project involves experts from INRA, and its goal is to propose
a proof of concept for a decision support tool allowing the formulation of well-
balanced products in terms of nutritional requirements with acceptable sensory
qualities for the consumer while using eco-friendly production processes.
Available data are mainly research data, collected during projects studying
real or model cheeses and focussing on specific parameters. Those projects in-
volved domain experts from different domains in food science (see [3] for details):
1. the production process of French hard cheeses. Different parameters were
measured during the production steps, e.g. the composition of the milk (i.e.
lipid, protein, lactose and water content) or the pH of the product during
each step.
2. the sensory perception during in-mouth food breakdown. The main attributes
for texture are Springiness, Firmness, Granularity, Hardness and Moisture,
while the attributes for taste are Taste intensity, Salty, Sour and Sweet.
3. the rheological properties of cheeses, e.g. The Young modulus.
4. the life cycle assessment for the production of stabilized micro-organisms.
Data Integration in Food Production 3
3 Data and Knowledge Representation Task
Each domain presented in the previous section has its specific concepts and
terms, and the domain experts from these different domains need to share the
same vocabulary in order to integrate available data of the NutriSensAl project.
Therefore an ontology named PO2 DG has been developed using Scenario 6 of
the NeON methodology [1], i.e. reusing, merging and re-engineering ontological
resources. It extends PO2 [4], a core ontology for process and observation, which
reuses BFO [5], IAO [6], OM [7], and naRyQ ontology [8]. PO2 DG reuses concepts
from GACS [9], AgroVoc [10] and NALT [11].
The core ontology PO2 and the domain ontology PO2 DG are both implemented
in OWL 2 [12]. The core ontology PO2 contains 90 classes, 122 properties and 15
individuals. The PO2 DG ontology contains 3475 classes, 122 properties and 6760
individuals.
Both ontologies are available on the the AgroPortal repository, http://
agroportal.lirmm.fr/ontologies/PO2 and http://agroportal.lirmm.fr/
ontologies/PO2_DG, and both are under the licence Creative Commons At-
tribution International 4.0 International (CC BY 4.0) [13].
A tool, PO2 VocabularyManager, was developed in order to help build PO2 DG
by first performing a syntactic search in a set of existing resources, then by as-
sisting the user in the concept creation task. This new tool reduces the time
necessary for building the domain ontology and allows a more efficient contribu-
tion of the domain experts in this phase of building the domain ontology.
Available data concerning the ongoing NutriSensAl project is now under
the process of integration into the PO2 DG dataset RDF repository available on
http://sonorus.agroparistech.fr:7200/.
The following steps were needed for data integration:
1. the knowledge engineer defined a set of EXCEL files structured according to
the PO2 DG ontology;
2. the domain experts were trained how to fill in these EXCEL files and a
guideline was written for the domain experts;
3. a script was written by the computer scientists in order to lift data from
EXCEL files into the RDF database.
Our experience shows that the second step is very difficult, error-prone and
time-consuming because of the complexity of the data. A second tool, PO2 DataMa-
nager, is under development and will be available soon. It will assist the domain
experts in integrating data in a more user-friendly way.
Fourteen projects are now stored in the PO2 DG dataset repository with 358
processes belonging to four distinct process types. 180 cheese samples are stored
in PO2 DG dataset.
4 Discussion
In [3] five competency questions concerning the NutriSensAl project are pre-
sented along with the corresponding SPARQL queries. Trying to answer cross-
4 L. Ibanescu et al.
domain questions shows that there was not enough data. When trying to es-
timate missing data using the available one, very important domain questions
arise: i) what method can be used for the estimation? ii) what available data
should be used?, and iii) what are ‘similar’ data?. Domain experts need to find
their answers to these questions and knowledge engineers can help.
The question what are ‘similar’ data? is also very challenging in computer sci-
ence and is investigated more and more in the context of linked data [14, 15]. The
need to define ‘similar’ products and ‘similar’ processes in NutriSensAl project
inspired in [16] the definition of a contextual identity link and an algorithm for its
detection in a knowledge base. Preliminary results on the PO2 DG dataset show
how to use the contextual identity link in order to predict missing observation
measures.
5 Conclusion
This paper presents the data representation task of the NutriSensAl project
consisting of the design of a domain ontology and the integration of available
data into an RDF repository using Web Semantic technologies. The preliminary
results illustrate how this semantic approach can be useful to estimate missing
data. Further work is to integrate more data into the PO2 DG dataset repository
and one of the next tasks is to explore how industrial data could be integrated
and if open data available on the LOD could be reused.
Another issue we would like to explore is to evaluate the FAIRness [17] of
the domain ontology and of the data stored in the RDF repository.
The data and knowledge representation and integration task of the NutriSen-
sAl project represents the first step and the backbone of the decision support tool
to be designed and implemented handling multi-criteria indicators. The aim of
the NutriSensAl project is to better understand and control the food production
process in order to formulate well-balanced products with a low environmental
impact.
References
1. Suárez-Figueroa, M.d.C., Gómez-Pérez, A., Fernández-López, M.: The NeOn
methodology for ontology engineering. In del Carmen Suárez-Figueroa, M., Gómez-
Pérez, A., Motta, E., Gangemi, A., eds.: Ontology Engineering in a Networked
World. Springer (2012) 9–34
2. : Linked Open Data. https://lod-cloud.net/
3. Pénicaud, C., Ibanescu, L., Allard, T., Fonseca, F., Dervaux, S., Perret, B.,
Guillemin, H., Buchin, S., Salles, C., Dibie, J., Guichard, E.: Relating transfor-
mation process, eco-design, composition and sensory quality in cheeses using PO2
ontology. International Dairy Journal (accepted for publication 2018)
4. Ibanescu, L., Dibie, J., Dervaux, S., Guichard, E., Raad, J.: PO2 - A Process and
Observation Ontology in Food Science. Application to Dairy Gels. In Garoufallou,
E., Coll, I.S., Stellato, A., Greenberg, J., eds.: Metadata and Semantics Research -
10th International Conference, MTSR 2016, Göttingen, Germany, November 22-25,
Data Integration in Food Production 5
2016, Proceedings. Volume 672 of Communications in Computer and Information
Science. (2016) 155–165
5. : Basic Formal Ontology (BFO). https://github.com/BFO-ontology/BFO
6. : Information Artifact Ontology (IAO). https://bioportal.bioontology.org/
ontologies/IAO
7. : Ontology of units of Measure (OM). https://github.com/HajoRijgersberg/OM
8. Buche, P., Dibie, J., Ibanescu, L., Soler, L.: Fuzzy web data tables integration
guided by an ontological and terminological resource. IEEE Trans. Knowl. Data
Eng. 25(4) (2013) 805–819
9. : Global Agricultural Concept Scheme (GACS). http://aims.fao.org/global_
agricultural_concept_scheme_gacs
10. : AGROVOC. http://aims.fao.org/vest-registry/vocabularies/agrovoc
11. : National Agricultural Library’s Agricultural Thesaurus (NALT). https://
agclass.nal.usda.gov/agt.shtml
12. : Web Ontology Language (OWL). https://www.w3.org/2001/sw/wiki/OWL
13. : Creative Commons Attribution International 4.0 International (CC BY 4.0).
https://creativecommons.org/licenses/by/4.0/
14. Halpin, H., Hayes, P.J., McCusker, J.P., McGuinness, D.L., Thompson, H.S.: When
owl: sameas isn’t the same: An analysis of identity in linked data. In Patel-
Schneider, P.F., Pan, Y., Hitzler, P., Mika, P., Zhang, L., Pan, J.Z., Horrocks,
I., Glimm, B., eds.: The Semantic Web - ISWC 2010 - 9th International Semantic
Web Conference, ISWC 2010, Shanghai, China, November 7-11, 2010, Revised Se-
lected Papers, Part I. Volume 6496 of Lecture Notes in Computer Science., Springer
(2010) 305–320
15. Beek, W., Raad, J., Wielemaker, J., van Harmelen, F.: sameas.cc: The closure
of 500m owl: sameas statements. In Gangemi, A., Navigli, R., Vidal, M., Hitzler,
P., Troncy, R., Hollink, L., Tordai, A., Alam, M., eds.: The Semantic Web - 15th
International Conference, ESWC 2018, Heraklion, Crete, Greece, June 3-7, 2018,
Proceedings. Volume 10843 of Lecture Notes in Computer Science., Springer (2018)
65–80
16. Raad, J., Pernelle, N., Saı̈s, F.: Detection of contextual identity links in a knowledge
base. In Corcho, Ó., Janowicz, K., Rizzo, G., Tiddi, I., Garijo, D., eds.: Proceedings
of the Knowledge Capture Conference, K-CAP 2017, Austin, TX, USA, December
4-6, 2017, ACM (2017) 8:1–8:8
17. : The FAIR data principles. https://www.force11.org/group/fairgroup/
fairprinciples