=Paper=
{{Paper
|id=Vol-2519/short1
|storemode=property
|title=Connecting Open Data and Sustainable Development Goals using a Semantic Knowledge Graph Approach
|pdfUrl=https://ceur-ws.org/Vol-2519/short1.pdf
|volume=Vol-2519
|authors=José Eduardo Eguiguren Palacios,Nelson Piedra
|dblpUrl=https://dblp.org/rec/conf/ontobras/PalaciosP19
}}
==Connecting Open Data and Sustainable Development Goals using a Semantic Knowledge Graph Approach==
https://ceur-ws.org/Vol-2519/short1.pdf
Connecting Open Data and Sustainable Development
Goals using a Semantic Knowledge Graph approach
José Eduardo Eguiguren1, Nelson Piedra1
1
Universidad Técnica Particular de Loja, Ecuador
{jeeguiguren, nopiedra}@utpl.edu.ec
Abstract. This paper aims to present an initiative to establish links and
relationships between Open Data (OD) and the Sustainable Development
Goals (SDG). OD published by various organizations using the CKAN
platform is highly dispersed and heterogeneous, making it harder to process
and leverage those vast amounts of information in their current state. This
paper approaches this opportunity by defining the steps that will allow the
construction of a semantic representation of data sets found in public access
data portals and each SDG using well-established ontologies and
vocabularies. Doing so will provide the necessary tools to link the many
resources containing data useful to the SDGs and relate them accordingly. Its
implementation is a work in progress.
1. Introduction
To achieve a better and more sustainable future for everyone, the United Nations
General Assembly stablished 17 Sustainable Development Goals1 with 169 targets to
work towards. Their objective is to provide a solution to all the global challenges we
face by 2030 (Hák, Janoušková, & Moldan, 2016). Amongst the many efforts that are
being done to achieve each goal, capturing data to monitor and measure the SDG targets
has become one of the most important (United Nations, n.d.). There are different
initiatives that encourage the collection of data that supports SDGs. The Global
Partnership for Sustainable Development Data was launched in 2015 in order to
“strengthen data-driven decision-making” to help accomplish the SDG (Adams, 2015).
To fully implement and monitor progress on the SDG, decision makers,
researchers, entrepreneurs and anyone interested need data and statistics that are open,
reliable, accessible, accurate, timely, sufficiently disaggregated, relevant, and easy to
use. Public access data portals are designed to collect and share open data. The
publication of open data to which anyone access promotes transparency and
collaboration. However, the primary obstacle presented by the nature of Open Data is
the high level of diversity found between different datasets. Because standards for
structuring data are vaguely used or often completely ignored, data portals end up with
heterogeneous sets of data that further complicate its re-use and re-purpose. These
portals can contain open data, at different levels of aggregation, related to the global
SDG indicators, and that can be used to discover, understand, or communicate patterns
and interrelationships between SDG and wealth of open data that are now available.
1 See https://www.un.org/sustainabledevelopment/sustainable-development-goals/
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
Harnessing the power of digital transformation through open data and semantic
technologies will be essential if we are to maximize the probability of implement and
monitoring progress on the SDG. In this work, the authors propose a way to link open
datasets with SDG indicators. The ultimate purpose is to develop the necessary tools for
open data consumers to globally discover data related to the SDG, deploy new open
data, and address these goals. By taking this opportunity, we can lay the foundation in
creating the aforementioned processes, and draw a parallel between the datasets
allocated in public access data portals and the SDG. Authors present a structure of Open
Data retrieved from CKAN2 repositories that are then represented semantically using
the vocabularies DCAT3 and Data cube4. The end goal is to link and relate them to the
various indicators defined in the SDG. Section 2 describes some principles of Linked
Data as well as concepts that will be used in this work. Then, section 3 defines the
approach used in order to achieve said goals as well as some implementations that have
been tested already. Finally, conclusions and future work are summarized in section 4.
2. Background
2.1. Linked Data
Tim Berners-Lee and his team in the World Wide Web Consortium (W3C) oversee the
process of designing the Semantic Web (SW) that proposes the next step for the
architecture of the Web. Its current goal is to provide a knowledge representation of LD
as well as increasing the amount of Web resources that are easily interpreted by
programs or agents (Alesso & Smith, 2006).
The Semantic Web defines the necessary practices and requirements to achieve a
Structured Web. One of these is the use of Linked Data (LD) which is defined by
(Bizer, 2009) as the “set of best practices to publish and connect structured data on the
web”. LD is also said to be the next step for Open Data where the information is
interconnected, and can be related to other groups of information (Piedra, Chicaiza,
López, & Caro, 2017).
LD’s principles can be utilized in any field where the main objective is to
centralize all information from various sources and stablish relationships that enrich the
data that can be found. One of these examples, that satisfy the mentioned criteria, is the
SmartLand-LD initiative (http://smartland.utpl.edu.ec) described in (Piedra & Suárez,
2018). One of its goals is to integrate every component that generates biologic, social,
economic and environmental information distributed in different networks. By doing
this they seek to provide a semantic structure in which its data will be exploited and
linked to the Sustainable Development Goals.
The benefits of LD can be seen very clearly when doing a recollection of
information across many different data sources. Such is the case for (Piedra, Tovar,
Colomo-Palacios, Lopez-Vargas, & Alexandra Chicaiza, 2014) where LD principles are
2
CKAN is a tool that easily allows the publishing and management of data sets. It is used globally by governments,
companies and organizations. Using CKAN allows us to maintain a free access to every resource without
restrictions, directly contributing to the main goal of Open Knowledge Foundation which is the organization
behind the administration of CKAN.
3 See https://www.w3.org/TR/vocab-dcat/
4 See https://www.w3.org/TR/vocab-data-cube/
applied in order to enhance the discoverability of OpenCourseWare (OCW) distributed
across many different universities. The solution provided includes the semantic
integration of the information found in OCW to allow for better filtering options
directed to end-users.
3. Proposed solution
The authors' solution arises from the following hypothesis: If there are so many Open
Data resources found in many different platforms comprising a multitude of knowledge
areas, there must be some of them that cover the same topics presented by each
Sustainable Development Goal. The sheer volume of data may provide information
useful to either track, measure or help each goal’s progress. Instead of having all those
sources of data separated in closed silos, we should integrate them and apply measures
that help us gather useful information about them from a different perspective. That
being the potential to link said resources to related SDG so that other people in need of
such information can find it easily. By using multiple platforms from different countries
and organizations we can cover a wide range of subjects that people from around the
world can use in their efforts to achieve each goal established by the UN.
Figure 1. Architecture defined to build a semantic integration from public
access data portals and link them with the SDG
In the next sections the authors describe the processes required to achieve our
goal. These processes are automated and included in a web application. The application
will then display the extracted and generated information in such a way that anyone can
filter, view it, and use it according to his or her interests. This information includes the
extracted data sets and the generated triples, as seen in Figure 1.
3.1. Extracting, transforming, linking and visualizing
The approach used in the process of building a semantic representation of OD found in
CKAN platforms is divided in three stages. First, we extract datasets from various
organizations that have implemented the CKAN platform to publish their data.
Afterwards we transform all data found to homogenize it as much as possible. Then,
continuing with the transformation, we use both DCAT and Data cube to build a
semantic representation of the cleansed OD. Finally, we store the generated triples and
proceed to the next phase.
The first challenge we encounter is finding a way to centralize all the available
data sources into a single data silo. Thankfully, we can leverage the help that the CKAN
(“CKAN,” 2018) platform brings in order to standardize the means of publishing data
and the extraction should be similar across the board.
Data visualization is one of the main end goals of this work where an
infrastructure is built on top of the semantic information to provide solutions for end
users to query OD, visualize and find similar datasets, integrate information to other
sources such as the SDG, and much more.
3.2. Data cube representation
The reasoning behind using Data cube to represent statistical data is that within each
resource, we may be able to find information that helps us describe the content of a data
set. To achieve this transformation, we have automated the extraction process of any
resource that has a CSV, XLS or XLSX file format. The values present in each file are
then iterated and automatically annotated using the Data cube vocabulary. However,
this process is currently experimental, and its results have not been used to relate the
data sets to each SDG.
3.3. DCAT representation
The DCAT vocabulary provides the backbone of the semantic structure for public
access data portals. Currently, after the data has been extracted automatically from the
CKAN sources, the relational data is then transformed to match the vocabulary with the
available metadata.
The transformation process has been automated using Apache Jenna5 in Java.
After all the data sets have been extracted and cleaned as seen in Figure 1, each one is
mapped accordingly using the DCAT vocabulary automatically. The result is a set of
triples that represent every extracted data set from each platform.
3.5 Transforming the SDG into a SKOS taxonomy
Once the data sets have been successfully transformed into a semantic knowledge
representation, the authors need a way to find relationships between them and the SDG.
Thus, the information present in each goal, target and indicator must be transformed
into a semantic representation as well. The challenge faced by the authors is that the
SDG are written in natural language. The following tool was found during research that
tackles this exact problem.
5 See https://jena.apache.org/
FRED is a tool whose purpose is to provide a machine reader for the Semantic
Web. Alani et al., (2016) states how this tool uses multiple Natural Language
Processing (NLP) components and unifies their result into a formal RDF/OWL graph.
In this work FRED was used as a mediator in charge of analyzing the content of each
goal, target and indicator from the SDG. The result provided by the tool is a formal
representation of the most relevant knowledge of the SDG elements. FRED also has a
key advantage that it automatically links the content found to already existing semantic
knowledge such as DBpedia6 resources.
The next step is to transform said graphs into an SKOS7 representation so that
we can stablish links with the resources from public data portals. The transformations
process involves using a set of queries that can extract the elements shown in Figure 2.
Each class is transformed into a SKOS concept that is then associated to a scheme based
on the goal, target or indicator. The relationships found can be established using the
related property of SKOS. If FRED found links to semantic web knowledge, we
associate them with the concepts using the DC Terms property subject8.
Figure 2. SDG Indicator 1.1.1 after being processed by FRED and the elements
used in the SKOS transformation
3.5. Linking Open Data with the SDG
At this point the authors have successfully created a semantic knowledge graph
representation out of Open Data and the SDGs. The final step was to find relationships
between them that could tell us if a dataset has something to do with one or more SDG.
This is done by analyzing the data found in every property from each dataset and
finding similarities with the concepts from the SDG representation.
6 See https://wiki.dbpedia.org/
7 See https://www.w3.org/TR/swbp-skos-core-spec/
8 See http://dublincore.org/specifications/dublin-core/dcmi-terms/2012-06-14/?v=terms#subject
The authors have chosen to use the SKOS vocabulary to represent the SDG for
one main reason. The vocabularies used to represent OD already implement SKOS in
their structure. This means that the authors can find similar concepts present in the SDG
and OD representations. SDG representation the authors can discover which goals are
data sets related to.
Our first approach involved analyzing the themes of each Data set and find
matches with the concept taxonomies generated from each SDG. In this first iteration,
we just analyzed the goals, excluding the targets and indicators to test the process with a
smaller subset of information. We created a semantic property called
“automaticallyAnnotatedSubject” that links any resource that has a match with an SDG.
From 42,924 extracted data sets, we managed to create 6,295 links between them and
the SDGs. Then the application is in charge of displaying which resources are related to
each SDG. Useful information can be found using filtering tools provided by the app.
4. Conclusions and Future Work
In this work we have presented how we can leverage the advantages provided by the
Semantic Web to allow for a better way to relate and distribute Open Data. A successful
transformation from the current state of public access data portals to a more descriptive
format using Linked Data can help overcome the numerous barriers that OD is currently
presenting. We can also use the resulting representation to link related data to other
fields such as the SDG. That way we can help to contribute in the global effort to
achieve said goals in a very meaningful manner.
The challenges that were resolved during this work, can be organized into the
following categories, a) technical issues: automatic extraction of open data from CKAN
portals, identifying errors and data gaps, conversion to linked open data formats,
maintaining Web interoperability standards and quality of the data, and creation of a
web tool to managing these functionalities. b) Legal and provenance issues: datasets
that lack explicit open licenses were assigned one based on the CKAN platform that
contains them. c) Interpretation of multiple languages is a challenge that seeks to
ensure no one is left behind; in this work, the prototype was applied only to datasets
described in the English language. However, the authors consider as future work to
meeting the needs of their users in multiple languages.
Future work includes finding more ways in which the Sustainable Development
Goals can be related to Open Data. This process involves analyzing the semantic graphs
that were produced using the previously discussed approaches. The authors have already
managed to create an infrastructure in which relationships can be stablished between the
SDG and OD. The following challenge involves the creation of useful information that
details those connections by testing this process using all the data extracted from the
previously mentioned open data portals.
Acknowledgement
The work has been funded and supported by the Universidad Técnica Particular de Loja
(UTPL).
References
Adams, B. (2015). SDG Indicators and Data : Who collects ? Who reports ? Who
benefts ? Policy Brief #9. Global Policy Watch, 1–8.
Alani, H., Gangemi, A., Presutti, V., Reforgiato Recupero, D., Giovanni Nuzzolese, A.,
Draicchio, F., & Mongiovì, M. (2016). Semantic Web Machine Reading with
FRED. Semantic Web, 0, 1–21. Retrieved from http://semantic-web-
journal.org/system/files/swj1379.pdf
Alesso, H. P., & Smith, C. F. (2006). Thinking on the Web: Berners-Lee, Godel and
Turing.
Bizer, C. (2009). The emerging web of linked data. IEEE Intelligent Systems, (5), 87-92.
CKAN. (2018). Retrieved from https://ckan.org/
Hák, T., Janoušková, S., & Moldan, B. (2016). Sustainable Development Goals: A need
for relevant indicators. Ecological Indicators, 60, 565–573.
https://doi.org/10.1016/j.ecolind.2015.08.003
Piedra, N., Chicaiza, J., López, J., & Caro, E. T. (2017). A rating system that open-data
repositories must satisfy to be considered OER: Reusing open data resources in
teaching. In Global Engineering Education Conference (EDUCON), 2017 IEEE
(pp. 1768–1777).
Piedra, N., & Suárez, J. P. (2018). SmartLand-LD: A Linked data approach for
integration of heterogeneous datasets to intelligent management of high
biodiversity territories. Advances in Intelligent Systems and Computing, 688, 207–
218. https://doi.org/10.1007/978-3-319-69341-5_19
Piedra, N., Tovar, E., Colomo-Palacios, R., Lopez-Vargas, J., & Alexandra Chicaiza, J.
(2014). Consuming and producing linked open data: the case of Opencourseware.
Program, 48(1), 16–40.
United Nations. (n.d.). United nations world data forum. Retrieved from
https://undataforum.org/