=Paper=
{{Paper
|id=Vol-3235/paper26
|storemode=property
|title=Building a Community-Based FAIR Metadata Schema for Brazilian Agriculture and Livestock Trading Data
|pdfUrl=https://ceur-ws.org/Vol-3235/paper26.pdf
|volume=Vol-3235
|authors=Filipi Miranda Soares,Fernando Elias Corrêa,Luis Ferreira Pires,Luiz Olavo Bonino da Silva Santos,Debora Pignatari Drucker,Kelly Rosa Braghetto,Dilvan de Abreu Moreira,Alexandre Cláudio Botazzo Delbem,Roberto Fray da Silva,Celso Oviedo da Silva Lopes,Antonio Mauro Saraiva
|dblpUrl=https://dblp.org/rec/conf/i-semantics/SoaresCPSDBMDSL22
}}
==Building a Community-Based FAIR Metadata Schema for Brazilian Agriculture and Livestock Trading Data
==
https://ceur-ws.org/Vol-3235/paper26.pdf
Building a Community-Based FAIR Metadata Schema for
Brazilian Agriculture and Livestock Trading Data
Filipi Miranda Soares1,2,8,∗,†, Fernando Elias Corrêa1,3,†, Luis Ferreira Pires2,
Luiz Olavo Bonino da Silva Santos2,4,†, Debora Pignatari Drucker1,5,†,
Kelly Rosa Braghetto1,6,†, Dilvan de Abreu Moreira1,7,†,
Alexandre Cláudio Botazzo Delbem1,7, Roberto Fray da Silva1,3,
Celso Oviedo da Silva Lopes1 and Antonio Mauro Saraiva1,3,8,†
1
University of Sao Paulo, Center for Artificial Intelligence (C4AI), Av. Prof. Lúcio Martins Rodrigues 370,
05508020 São Paulo SP Brazil
2
University of Twente, Faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS),
Drienerlolaan 5, 7522 ME Enschede Netherlands
3
University of Sao Paulo, Institute of Advanced Studies, R. do Anfiteatro 513, 05508060 São Paulo SP Brazil
4
Leiden University Medical Center (LUMC), Albinusdreef 2, 2333 ZA Leiden Netherlands
5
Embrapa Digital Agriculture, Av. Dr. André Tosello 209, 13083-886 Campinas SP Brazil
6 University of Sao Paulo, Institute of Mathematics and Statistics, R. do Matão 1010, 05508090 São Paulo SP Brazil
7 University of Sao Paulo, Institute of Mathematics and Computer Sciences (ICMC), Av. Trab. São Carlense 400, 13566590,
São Carlos SP Brazil
8 University of Sao Paulo, Polytechnic School, Av. Prof. Luciano Gualberto 380, 05508010 São Paulo SP Brazil
Abstract
In this paper, we discuss how we are using metadata schemas and controlled vocabularies to
improve interoperability between Brazilian agriculture and livestock trading data providers. A
new metadata schema is being created based on a community-based approach. This method
relies on knowledge from specialists to define a list of relevant metadata properties for a given
domain. In the first step of the research, we extracted metadata from three datasets maintained
by three Brazilian public institutions: the Center for Advanced Studies in Applied Economics
(Cepea), the Institute of Applied Economic Research (Ipea), and The National Supply
Company (Conab). The extracted metadata were the input to the definition of a list of 15
potential metadata properties that specialists are validating.
Keywords 1
Agriculture, livestock, metadata, trading data, FAIR data principles
1. Introduction
The interoperability of information systems is largely affected by the availability of standards that
define common vocabularies and procedures. The large variety of incompatible data practices hinders
automated data exchange. Especially in agriculture, interoperability poses a significant challenge due
to the multidisciplinary and interdisciplinary nature of the field.
Digital agriculture is characterized by the massive use of technology in crop and livestock
production, either by using more efficient and intelligent equipment such as mobile technologies,
remote sensing services, and robotics [1], or by the use of data and software technologies, such as
databases, adaptive systems, Artificial Intelligence, Big Data, Internet of Things, Virtual and
Augmented Reality [1]. Although there are some advances in digital agriculture, more efforts in this
SEMANTICS 2022 EU: 18th International Conference on Semantic Systems, September 13-15, 2022, Vienna, Austria
EMAIL: filipisoares@usp.br (A. 1); fecorrea@usp.br (A. 2)
ORCID: 0000-0002-0674-7960 (A. 1); 0000-0002-2744-6104 (A. 2); 0000-0001-7432-7653 (A. 3); 0000-0002-1164-1351 (A. 4); 0000-0003-
4177-1322 (A. 5); 0000-0001-6218-6849 (A. 6); 0000-0002-4801-2225 (A. 7); 0000-0003-1810-1742 (A. 9); 0000-0002-9792-0553 (A. 10);
0000-0002-7004-6042; 0000-0003-2283-1123 (A. 11)
©️ 2020 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
direction are necessary as agriculture is still considered one of the least digitized productive sectors in
the world [2].
In this context, the agribusiness sector, which deals with trading, among other things, is
demonstrating increasing interest in digital agriculture and is capturing huge financial investments [3].
Thus, an increasing number of scientists worldwide have been committed to collecting, manipulating,
and storing agribusiness data, especially data from Brazilian agribusiness, given the high Brazilian
agricultural production capacity and the importance of agriculture to the country. Therefore, we believe
metadata schemas that can cope with the complexity of agricultural data and have the potential to be
standardized can significantly change how we produce and share information on agriculture, especially
in Agribusiness.
In this paper, we present an initial effort to build a metadata schema for agricultural trading data to
allow data interoperability between three of the main Brazilian data providers on agricultural
commodities: the Center for Advanced Studies in Applied Economics (Cepea2), the Institute of Applied
Economic Research (Ipea3), and the National Supply Company (Conab4). We address the challenges
and issues faced and the strategies we are using to develop the semantic structure of the metadata
elements.
The paper is further structured as follows: Section 2 gives the background of our work, Section 3
discusses the methodology we have been following, Section 4 presents our results so far, and Section 5
gives our conclusions.
2. Background: metadata schemas and metadata records
Metadata is data about data [4][5]. This definition, however, does not say much since it does not
define which aspects of the data are represented in the metadata. As pointed out by Coyle [6], metadata
are more complex than just data: they are human-made artifacts created with a purpose and to perform
a function in data description. Mayernik [7] presented a literature review that included many definitions
for metadata. Among them is the definition of Greenberg [4], which considered metadata as “data
attributes that describe, provide context, indicate the quality, or document other object (or data)
characteristics”, and the definition of Smiraglia [8], which defined metadata as “structured descriptors
of information resources, designed to promote information retrieval”. From these definitions, we can
consider metadata as attributes or properties that describe characteristics of data resources for
information retrieval and to allow the correct interpretation and use of data, amongst other purposes.
Metadata and data are often registered in digital records as property-value pairs. A set of metadata
is used to represent a subject, which is an entity from the real world that one wants to describe. Thereby,
metadata can be represented as triples — subject-predicate-object [5]. A set of triples forms a graph,
where the subjects and objects are the nodes, and the predicates are the edges.
Metadata is usually part of a metadata schema. The metadata properties in a metadata schema have
two components [9]: semantic, i.e., the definition of the meaning of the term that can be used as a
metadata field, and syntax, i.e., the recommendation or specification of the format of data (codification)
that should be assigned to a metadata field.
With the advances of the Semantic Web and linked data, new technologies have emerged to improve
the use of metadata, especially by machines. One of the most prominent technologies is the Resource
Description Framework (RDF), which associates Unique Resource Identifiers (URI) to terms assigned
as metadata elements. To be used in RDF parameters, namespaces have been assigned to metadata
schemas to preserve the context of the metadata elements in the multiple environments in which they
can be used.
In the Agriculture domain, the Food and Agriculture Organization of the United Nations (FAO) has
developed the Agricultural Metadata Element Set (AGMES), which is a metadata schema that describes
a variety of digital resources for agriculture [10]. Many other communities have been working on
2
Available from: https://www.cepea.esalq.usp.br/br, last accessed 2022/07/04.
3
Available from: http://www.ipeadata.gov.br/Default.aspx, last accessed 2022/07/04.
4
Available from: https://sisdep.conab.gov.br/precosiagroweb/, last accessed 2022/07/04.
developing ontologies and other kinds of controlled vocabularies for agriculture, with more than a
hundred available on AgroPortal5.
3. Methodology
We started by analyzing three of the largest Brazilian agricultural commodities datasets to
understand which metadata they apply to their data. These three datasets were selected a priori based
on the solid scientific reputation and recognized contribution to the Brazilian Community of the
organizations behind these datasets, namely Cepea, Conab, and Ipea. These datasets contain essential
statistical data from Brazil's agricultural trades. Within these data, we decided to set the scope on time-
series data of market prices since all three datasets contain this kind of data.
Through this consultation, we extracted descriptive information that was available along with the
data on the websites of these data providers. A specialist then validated the information obtained.
From the analysis carried out, three new processes were identified: (i) definition of the list of
common descriptive words used to fulfill the metadata fields, such as product names (e.g. cotton, cattle,
rice, soybeans), units of measure (e.g. box of 15kg), locations, among others; (ii) harmonization of the
data for the product name metadata field, by adopting the controlled vocabulary Agrotermos, an
agriculture thesaurus that has the largest base of agriculture terminology in Brazilian Portuguese, and
is a terminology provider for Agrovoc, the FAO multilingual thesaurus; (iii) based on the most
representative descriptors common within three datasets, selection of the first list of 15 potential
metadata properties for agricultural commodities. In this step, we gave names and definitions to these
metadata properties.
We are now querying the community of practice about this first set of metadata properties to validate
the terms and definitions and suggest other terms that could be significant to the domain. For this
purpose, specialists from nine Brazilian organizations participating in the 5th Brazilian Open
Government Action Plan6 were asked to annotate sample datasets with the 15 metadata properties. In
addition, they will answer a System Usability Scale (SUS) questionnaire adapted from the model
presented by Tullis & Albert [11], originally developed to evaluate information systems’ interfaces.
4. Results
The complete list of metadata is under review, so the first version of the metadata schema is not
available on RDF or any other Semantic Web format yet. The 15 metadata properties we used to
illustrate our approach and their definitions are the following:
• Product group: types of products. The best practice is to use a controlled vocabulary such as
Agrotermos. Examples: grain, vegetable, meat.
• Theme: main topic investigated in the statistical operation. The best practice is to use a
controlled vocabulary. Examples: regional average price, indicator, index, cost, technical or
zootechnical coefficient.
• Product name: name of the agricultural or livestock product. The best practice is to use a
controlled vocabulary such as Agrotermos. Examples: soy, corn, cattle.
• Verbatim name: natural language name given to the data series in the original dataset.
• Publisher: entity responsible for making the resource available7.
• Creator: entity responsible for creating the resource8.
• References: related resources that are referenced, cited, or otherwise pointed to by the described
resource9.
• Data type: kind of the data that make up the record. The best practice is to use a controlled
vocabulary. Examples: multivariate, univariate, synthetic, sequential, time series, text.
5
Available from: http://agroportal.lirmm.fr/, last accessed 2022/09/02.
6
Available from: https://wiki.rnp.br/pages/viewpage.action?pageId=155657461, last accessed 2022/07/04.
7
Available from: http://purl.org/dc/terms/publisher, last accessed 2022/07/04.
8
Available from: http://purl.org/dc/terms/creator, last accessed 2022/07/04.
9
Available from: http://purl.org/dc/terms/references, last accessed 2022/07/04.
• Frequency: temporal frequency of data publication. Examples: daily, monthly, bimonthly,
yearly.
• Unity: measure or quantity of the product. Examples: 'arroba' (Brazilian unit of weight,
equivalent to 15kg, represented by the symbol @); a bag of 30kg.
• Location: place or region to which the data series refers. The best practice is to use geographic
coordinates (decimal latitude and longitude) or insert names according to Geonames10.
• Start of reference period: date of publication of the first dataset of a data series. The best-
recommended practice is to adopt date encoding schemes such as ISO 8601 or ABNT NBR 5892.
• End of reference period: date of publication of the last dataset of a data series. The best-
recommended practice is to adopt date encoding schemes such as ISO 8601 or ABNT NBR 5892.
In the case of a current series, this field must be left empty.
• License: A legal document that defines the policy to do something with the resource11. The
best-recommended practice is to adopt a URI to indicate the license.
• Methodology: summary of the methods used to generate the dataset. The best practice is to
indicate the resource URI published in an open Web access format12.
The Product group property has been defined to group data records into broader product categories,
while the Product name property specifies the subgroup of the product. In a metadata record about
carrots, for example, the Product group would be vegetable, and the product name would be carrot.
We are using the Agrotermos thesaurus to standardize data for product group and name properties.
However, although Agrotermos has a huge base of agriculture terminology, some important terms from
the trade’s domain are still missing, such as, for example, 'boi gordo' (Portuguese for fed cattle), which
is a very popular term to designate a specific kind of cattle in the Brazilian trade market. Therefore, we
are working with the Agrotermos Developing Team13 at the Brazilian Agricultural Research
Corporation (Embrapa) to include missing terms in Agrotermos.
Besides the controlled vocabulary fields, the verbatim name property allows the name given in the
original dataset to be preserved. The Theme property indicates the kind of trade indicators, which is
deeply related to the methodology property since each kind of trade indicator follows a possibly
different method.
Publisher and Creator were imported from the Dublin Core schema. The Publisher property can be
used to refer to the institution that published the datasets (e.g., Cepea, Ipea, Conab). In contrast, the
Creator property refers more specifically to the person or department within those institutions
responsible for generating the datasets. The References property was also imported from Dublin Core
and gives the link to the address of the dataset on the Web. The License property was another Dublin
Core term we incorporated into our metadata schema. It describes the license of use of the dataset given
by the data provider.
Other properties, i.e., Data type, Frequency, Unity, Location, Start of reference period, and End of
reference period are often variants of terms already present in generic metadata schemas, such as
Schema.org14, Wikidata properties15, and the Data Catalog Vocabulary (DCAT16), so in a later step of
the research, we will identify and sort out which of these generic metadata vocabularies should be
referred to in our metadata schema.
The metadata schema presented in this paper has been named Agriculture and Livestock Metadata
Elements Set (Almes), or Almes Core (short name). The name hints at a broader scope than the metadata
schema presented in this paper, which only focuses on trading data. This is because we intend to extend
the metadata schema to cover other sub-disciplines of agriculture and livestock in the upcoming years
of the project.
10
Available from: http://www.geonames.org/, last accessed 2022/09/03.
11
Available from: http://purl.org/dc/terms/license, last accessed 2022/07/04.
12
Available from: https://metadados.ibge.gov.br/consulta/glossario/estatistico, last accessed 2022/07/04.
13
The full name of the Team in Portuguese: Comissão Permanente de Trabalho em Vocabulários Controlados, Agroterminologias e
Agrossemântica da Embrapa (GTermos).
14
Available from: https://schema.org/, last accessed 2022/07/04.
15
Available from: https://www.wikidata.org/wiki/Wikidata:List_of_properties, last accessed 2022/07/04.
16
Available from: https://www.w3.org/TR/vocab-dcat-2/, last accessed 2022/07/04.
5. Final considerations
We expect the metadata schema we are developing will improve the interoperability within
agriculture and livestock trading data providers in Brazil and that it will also be useful for similar
purposes in other countries after being extended. We also expect that the methods to make our metadata
schema FAIR applied in this research can be reused by other initiatives with similar purposes.
In the following steps, we will review the metadata properties based on suggestions from the
community of practice. After that, we move forward to publish the metadata schema and all the
documentation needed so that the community can use it. The FAIR principles will be applied in the
publication process to make the metadata schema findable on the web, accessible without restrictions,
interoperable with other metadata schemas, and reusable by humans and machines [12]. The FAIR Data
Points [13] will be implemented to enforce the FAIRness of the data repositories containing harmonized
datasets described with the Almes Core properties.
Agriculture is a vast subject, so we expect to address other sub-disciplines to create other metadata
sets in the future. Given the country's considerable biodiversity and climate conditions, Brazilian
agriculture has many peculiarities, so we expect to develop controlled vocabularies in Portuguese (with
translations to English) that can better suit the needs of the Brazilian agriculture and livestock data
community.
6. Acknowledgments
FMS thanks the São Paulo Research Foundation (FAPESP) for the research grants n. 2021/15125-0
and 2022/08385-8. The authors thank the Center for Artificial Intelligence (C4AI), a joint initiative of
USP, IBM and FAPESP (process n. 2019/07665-4), for the technical and/or financial support to their
research. AMS thanks the Brazilian National Council for Scientific and Technological Development
(CNPq) for the research grant n. 312605/2018-8.
7. References
[1] Shamin, A., Frolova, O., Makarychev, V., Yashkova, N., Kornilova, L., Akimov, A. Digital
transformation of agricultural industry. IOP Conference Series: Earth and Environmental Science,
346(1), (2019). https://doi.org/10.1088/1755-1315/346/1/012029
[2] Saraiva, A. M., Costa, W. F., Xavier, F., Albertini, B. de C., Pfeifer, R. A. C., Júnior, M. A.,
Simplício, A. K. V. Dados na Agricultura Digital: ciclo, padronização, qualidade,
compartilhamento e segurança. In Daniel Marçal Queiroz, D. S. M. Valente, F. de A. C. Pinto, &
A. Borém (eds.).: Agricultura Digital. UFV, Viçosa, pp. 308–325 (2020).
[3] Duncan, E., Rotz, S., Magnan, A., Bronson, K. Disciplining land through data: The role of
agricultural technologies in farmland assetisation. Sociologia Ruralis 62(2), 231–249 (2022).
https://doi.org/10.1111/SORU.12369
[4] Greenberg, J. Understanding Metadata and Metadata Schemes. Cataloging & Classification
Quarterly, 40(3-4), 17–36 (2005). https://doi.org/10.1300/J104V40N03_02
[5] Pomerantz, J. Metadata. The MIT Press, Cambridge (2015).
[6] Coyle, K. Understanding Metadata and Its Purpose. The Journal of Academic Librarianship, 31(2),
160–163 (2005). https://doi.org/10.1016/J.ACALIB.2004.12.010
[7] Mayernik, M. S. Metadata. Knowledge Organization, 47(8), 696–713 (2020).
https://doi.org/10.5771/0943-7444-2020-8-696
[8] Smiraglia, R. P. Metadata: a cataloger’s primer. Routledge, Abingdon-on-Thames (2005).
[9] Chan, L. M., Zeng, M. L. Metadata Interoperability and Standardization - A Study of Methodology
Part I. D-Lib Magazine, 12(6), (2006). https://doi.org/10.1045/june2006-chan
[10] Food and Agriculture Organization of the United Nations (FAO). AgMES 1.1 Namespace
Specification (2010). http://aims.fao.org/standards/agmes/namespace-specification, last accessed
2022/07/04.
[11] Tullis, T., Albert, B. Measuring the User Experience. 2nd ed. Elsevier, Amsterdam, (2013).
[12] Wilkinson, M. D., Dumontier, M., Aalbersberg, Ij. J., Appleton, G., Axton, M., Baak, A.,
Blomberg, N., Boiten, J. W., da Silva Santos, L. B., Bourne, P. E., Bouwman, J., Brookes, A. J.,
Clark, T., Crosas, M., Dillo, I., Dumon, O., Edmunds, S., Evelo, C. T., Finkers, R., … Mons, B.
The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3,
1–9 (2016). https://doi.org/10.1038/sdata.2016.18
[13] Bonino, L., Kaliyaperumal, R., Kees Burger, A., Kuzniar, A. Gavai. FAIR Data Point Specification
(2021), https://github.com/FAIRDataTeam/FAIRDataPoint/wiki, last accessed 2022/07/04.