A Linked Data-based Service for Integrating Heterogeneous Data
Sources in Smart Cities
João Gabriel Almeida, Jorge Silva, Thais Batista
a
and Everton Cavalcante
b
Federal University of Rio Grande do Norte, Natal, Brazil
Keywords:
Heterogeneous Data, Data Integration, Linked Data, NGSI-LD, Smart Cities.
Abstract:
The evolution and development of new technological solutions for smart cities has significantly grown in recent
years. The smart city scenario encompasses larges amount of data from several devices and applications. This
raises challenges related to data interoperability, including information sharing, receiving data from multiple
sources (Web services, files, systems, etc.), and making them available at underlying smart city application
development platforms. This paper presents Aquedücte, a service that converts data from external sources and
files to the NGSI-LD protocol, enabling their use by applications relying on an NGSI-LD-based middleware.
This paper describes the Aquedücte methodology used to: (i) extract data from heterogeneous data sources, (ii)
enrich them according to the NGSI-LD data format using Linked Data along with ontologies, and (iii) publish
them into an NSGI-LD-based middleware. The use of Aquedücte is also described in a real-world smart city
scenario.
1 INTRODUCTION
In the context of smart cities, a large amount of het-
erogeneous data is generated, stored (in various for-
mats), and exchanged through different communica-
tion protocols. These data sometimes are not readily
usable due to the heterogeneity of data types and for-
mats from vertical silos produced by smart city sys-
tems (d’Aquin et al., 2015). Making smart cities a re-
ality involves addressing interoperability at data level
towards reusing and sharing data. Data interoperabil-
ity can be achieved through the adoption of a stan-
dardized semantic-based data model that unifies the
format of data, provides a common meaning to them,
and allows for complex reasoning.
Recently, the combination of the Resource De-
scription Framework (RDF)
1
, the W3C Web Ontol-
ogy Language (OWL)
2
, and Linked data has been
considered the reference practice for sharing and pub-
lishing structured data on the Web (Bizer et al., 2009;
Consoli et al., 2017). Linked data allows integrat-
ing data into a common, browsable, accessible graph,
thereby allowing for the use of data across different
domains. It has been effective in many cases when in-
a
https://orcid.org/0000-0003-3558-1450
b
https://orcid.org/0000-0002-2475-5075
1
https://www.w3.org/TR/rdf-schema
2
https://www.w3.org/OWL/
formation from distinct sources must be put together
in a generic way and made available for a variety
of applications. By linking data, correlations can be
quickly understood. In the smart city context, an ex-
change protocol named Next Generation Service In-
terfaces - Linked Data (NGSI-LD)
3
has been recently
proposed to comply with the Linked Data concept.
NGSI-LD defines an information model based on en-
tities, relationships, and properties upon RDF, all of
them being semantically represented as concepts of
ontologies to ensure homogeneity among data from
different sources and contexts.
Even though a standardized data model eases data
integration from multiple heterogeneous sources, it is
necessary to define an approach that receives such
data and converts them to a target semantic-based
standardized data model. Aiming at tackling such a
challenge, this paper presents Aquedücte, a service
that imports data from different sources (third-party
Web services or files), converts them to an NGSI-LD-
based model, and makes them available to an under-
lying NGSI-LD-based middleware. Therefore, smart
city applications can be built upon that middleware
and hence exploit the available shared data.
This paper is structured as follows. Section 2
presents a background about Linked Data and the
3
https://bit.ly/2ORxbEV
Almeida, J., Silva, J., Batista, T. and Cavalcante, E.
A Linked Data-based Service for Integrating Heterogeneous Data Sources in Smart Cities.
DOI: 10.5220/0009422802050212
In Proceedings of the 22nd International Conference on Enterprise Information Systems (ICEIS 2020) - Volume 1, pages 205-212
ISBN: 978-989-758-423-7
Copyright
c
 2020 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
205

Figure 1: The NGSI-LD Information Model.
NGSI-LD data protocol. Section 3 describes the
Aquedücte architecture and its components. Section 4
describes a proof of concept regarding a real-world
smart city scenario. Section 5 briefly discusses about
related work. Section 6 contains final remarks.
2 BACKGROUND
Linked Data (Heath and Bizer, 2011) is a method
to publish interlinked data using standardized Web
technologies such as HTTP, RDF, and URIs aiming
at building complex information by aggregating sim-
pler information units. Linked Data allows for large-
scale integration of data in the Web and reasoning
about them. Using RDF, information is represented as
a triple <subject, predicate, object> in which an ob-
ject can be another subject, thus allowing interlinking
with each other. Such a linking structure forms a di-
rected labeled graph in which edges represent named
links between two resources, which are represented
by graph nodes.
The ETSI Industry Specification Group for Cross-
Cutting Context Information Management (ISG CIM)
has recently proposed the NGSI-LD specification,
which encompasses an information model with se-
mantic aspects related to Linked Data and ontolo-
gies. The main constructs of the NGSI-LD informa-
tion model are entities, relationships, and properties.
An entity represents a real-world object such as a
building or a person. A relationship associates enti-
ties, e.g., a person working in a building. A property
associates values with elements, such as identifying
the person name as Alice. Figure 1 depicts the main
elements of the NGSI-LD information model.
NGSI-LD entities are represented using JSON-
LD
4
, an extension of the JSON format now tailored
for Linked Data. The JSON-LD aims to serialize en-
tity data in a simple, effective way instead of using
RDF triples as commonly adopted in Linked Data.
4
https://json-ld.org/
Figure 2: Aquedücte Architecture.
Such an approach is interesting since it takes advan-
tage of the well-known JSON format, thus minimiz-
ing possible compatibility issues (Lanthaler, 2013).
3 THE AQUEDüCTE SERVICE
Aquedücte is a service aimed at standardizing data
from external Web sources and files while taking ad-
vantage of the facilities and resources provided by
the NGSI-LD protocol. The main concern is fostering
integration among heterogeneous data by enabling
them to be imported to any middleware infrastruc-
ture that works with NGSI-LD. Section 3.1 describes
the Aquedücte architecture. Section 3.2 presents some
implementation details and used technologies.
3.1 Architecture
The Aquedücte architecture is composed of two main
elements, namely (i) the Aquedücte UI for interacting
with users and (ii) the Aquedücte Service that imple-
ments its main functionalities. Figure 2 provides an
overview of the Aquedücte architecture.
The Aquedücte UI consists of a user-friendly in-
terface for loading and extracting data from RESTful
Web services or files. It consists of two components,
API Loader and File Loader: the former loads data
from Web services in the JSON format, whereas the
latter handles data in different file formats. This high-
level interface allows extracting, filtering, and con-
verting data to the NGSI-LD protocol.
The Aquedücte Service provides a RESTful in-
terface and services for NGSI-LD data extraction
and conversion operations, besides a non-relational
ICEIS 2020 - 22nd International Conference on Enterprise Information Systems
206

database to store import setups created by users. That
import setups consist of the parameters that were used
to perform data extraction, filtering and importing.
The use of the Aquedücte Service requires client
authentication provided by the Authenticator compo-
nent, which can use a third-party security service.
The main functionalities of the Aquedücte Service are
realized by four main components, namely (i) Files
and APIs Processor, (ii) Import Setup Manager, (iii)
NGSI-LD Converter, and (iv) Data Importer.
The Files and APIs Processor component is re-
sponsible for filtering and extracting data through
two different approaches, Context and Common. As
means of standardizing extracted data and easing data
analysis and queries with NGSI-LD, the Context ap-
proach takes advantage of LGeoSIM, a semantic in-
formation model for smart cities (Rocha et al., 2019).
LGeoSIM was chosen as information model as it is
able to address data heterogeneity while considering
georeferenced information, which can be quite use-
ful for smart city applications. LGeoSIM relies on the
ability of defining different layers, each one repre-
senting a set of related elements with georeferenced
information. Furthermore, LGeoSIM is grounded on
the NGSI-LD specification to allow for Linked Data
along with ontologies.
The Common approach concerns extracting, filter-
ing, and importing data without providing standard-
ization through a specific context source. This ap-
proach is valid for cases in which a given data set to
be manipulated does not require standardization via a
context source from the end-user’s point view. There-
fore, only using the NGSI-LD protocol would be suf-
ficient.
The Import Setup Manager component aims to
manage user settings to easing data import. Therefore,
users can select one of the registered setups to import
data, thus avoiding setting again previously defined
parameters.
The NGSI-LD Converter component is responsi-
ble for converting data either from external Web ser-
vices or files with different structures. This compo-
nent handles both common properties (expressed in
numeric, text, and collection types) and geolocation
properties (expressed in the GeoJSON format
5
) and
converts data sets structured as key-value pairs to the
NGSI-LD protocol.
Listing 1 shows an example of entity data com-
plying with the NGSI-LD protocol. id and type are
mandatory properties: the former is a unique iden-
tifier for the entity and the latter indicates the type
represented by the entity. Other properties follow
the NGSI-LD standard. In this example, the loca-
5
https://geojson.org/
tion property is of the GeoProperty type and has a
GeoJSON value representing geographic information
about the entity.
{
" id ": "ur n : ngsi -ld : Pa rk i ng : D ow nt ow n1 " ,
" ty pe " : " Pa rk ing " ,
" na me " : {
" ty pe " : " Pr op er ty " ,
" va lu e ": " Do wn t o w n One "
},
" t ot a l S po tN um b e r ": {
" ty pe " : " Pr op er ty " ,
" va lu e ": 200
},
" l oc at i on ": {
" ty pe " : " Ge o P r o p e r ty ",
" va lu e ": {
" ty pe " : " Po i nt " ,
" c oo rd in at es ": [ -8.5 , 41 .2]
}
},
" @ co nt e xt ": [
" ht tp :// uri . et s i . or g / ngsi - ld / v1 /
ngsi - ld - core - c ont ex t . j so nl d " ,
" ht tp :// e xa m pl e . o rg / ngsi - l d /
pa rk i ng . j so n ld "
]
}
Listing 1: Example of NGSI-LD compliant data.
The @context field contains links to ontology files
that define a data vocabulary for an entity as means
of ensuring semantic consistency. For instance, the
@context field could contain a link pointing to a data
vocabulary about vehicles with fields such as name
and engine. Therefore, any entity using this vocab-
ulary must contain at least one of these fields. It is
worth highlighting that validating syntactic and se-
mantic information about the entity against the data
vocabulary is out of the scope of Aquedücte.
Data Importer is a core component in the Aque-
dücte architecture. This component is responsible for
importing NGSI-LD data sets to an application or
middleware able to handle such a protocol. Data im-
port takes place through a RESTful communication
between the Aquedücte’s Data Importer component
and an external persistence service.
3.2 Implementation
Aquedücte was developed by following a RESTful
service-oriented architecture. The Aquedücte UI is a
front-end developed with Vuetify
6
, a Javascript-based
library that provides developers with ready-to-use UI
components. This library also comes with two-way
6
https://vuetifyjs.com/
A Linked Data-based Service for Integrating Heterogeneous Data Sources in Smart Cities
207

data binding or reactive approach, which enables de-
clared JavaScript variables to be synchronized with
any changes that may occur in the Document Object
Model (DOM) or UI.
The Aquedücte Service is a back-end developed
atop Spring Framework
7
, one of the most widely used
Java Web frameworks. The Spring Boot module
8
was
used to implement the Aquedücte Service to ease ini-
tial project configuration through the use of depen-
dencies. The Spring Data module
9
was used to persist
data at MongoDB, a non-relational database chosen
to the Aquedücte Service implementation. MongoDB
is used by the Import Setup Manager component as
means of storing import setups, which can be further
created and queried by users.
The NGSI-LD Converter component consists of
RESTful Web services for each approach for con-
verting data to the NGSI-LD protocol, namely Com-
mon and Context (see Section 3.1). In the Common
approach, the respective service receives an object
complying with NGSI-LD as payload, as shown in
Listing 2. Such a payload is composed of two fields,
geoLocationConfig and dataContentForNGSILDCon-
version. The former field constitutes the configura-
tions defined by the user to convert non-GeoJSON ge-
olocation data to the GeoProperty type, which has a
GeoJSON value. The latter field contains the data to
be converted to the NGSI-LD protocol.
{
" g eo L oc at io n Co nf ig ": [
{
" key ": " lo ca ti on ",
" t yp e O f Se le ct i o n ": " st ri ng " ,
" i nv er tC o o r d s ": true ,
" d el im it er ": "," ,
" t yp e G e ol oc at i o n ": " Po int "
}] ,
" d a ta Co n te n t F or N GS IL D Co n v e rs i o n ": [
{
" l oc at i on ": " -5 .15 9 421 90 ,
-37 .36 057 650 " ,
" c re D at e " : "19 96 -11 -12 " ,
" ci ty " : " Na t al " ,
" s ch oo lN am e " : " A th en eu " ,
" si ze " : " Big "
}, ...
]
}
Listing 2: Example of payload to convert data to NGSI-LD
through the Common approach.
The geoLocationConfig field has some attributes to
help converting geographic data:
7
https://spring.io
8
https://spring.io/projects/spring-boot
9
https://spring.io/projects/spring-data
• key stores the name of the field to be converted to
the GeoJSON format;
• typeOfSelection stores the type of field (string,
collection or GeoJSON) to be converted to Geo-
JSON;
• invertCoords is a Boolean attribute that indicates
if the order of latitude and longitude values need
to be inverted when the typeOfSelection attribute
is different from GeoJSON;
• delimiter consists of an attribute that stores a de-
limiter (pipe, comma or any other) when the type-
OfSelection attribute is of the string type;
• typeGeolocation stores a GeoJSON data type
(point, polygon), being relevant only when the
typeOfSelection attribute is different from Geo-
JSON as data formatted in GeoJSON already
comes with this attribute type.
The Context approach differs from the Common ap-
proach since the payload received by the service con-
tains two additional properties as means of maintain-
ing semantic consistency. Listing 3 shows how such a
payload is structured.
The contextLink property represents a JSON-LD
context file selected by the user with a data vocab-
ulary, which is used to standardize data from differ-
ent sources. This standardization is done through a
matching process in which the properties available at
the context file loaded by the user will be matched
with the ones derived from data extracted available at
external sources. This mapping generates the match-
ingConfigContent property with the following fields:
• contextName receives the property/attribute of the
loaded context file;
• foreignProperty receives the property/attribute
from data extracted from external sources;
• isLocation is a Boolean field that indicates if a
certain property derived from the extracted data
should be converted to GeoJSON, following the
settings defined by the geoLocationConfig field;
• geoLocationConfig has attributes to help convert-
ing geographic data.
4 PROOF OF CONCEPT
The Aquedücte service is currently integrated to
Smart Geo Layers (Souza et al., 2018), a smart city
middleware platform aimed to (i) integrate data
provided by heterogeneous sources, (ii) support data
correlation with geographic information, and (iii)
provide functionalities such as data aggregation,
visualization, querying, and analysis. In terms of se-
curity, Aquedücte uses the authentication and persis-
ICEIS 2020 - 22nd International Conference on Enterprise Information Systems
208

{
" c on te xt Li nk ":
" ht tp s : // u rl . co m / ngsi - ld / ed uc at io n /
sc hoo l / Sc ho o l _ Co nt ex t . js on ld " ,
" m a t c hi ng C on fi g Co nt e n t ": [
{
" c on te xt Na me ": " de sc ri pt io n " ,
" f or e i g nP ro pe r t y ": " s ch oo lN am e " ,
" i sL oc at io n " : f a lse ,
" g eo L oc at io n Co nf ig ": {}
},
{
" c on te xt Na me ": " city " ,
" f or e i g nP ro pe r t y ": " s ch oo lC it y " ,
" i sL oc at io n " : f a lse ,
" g eo L oc at io n Co nf ig ": {}
},
{
" c on te xt Na me ": " lo ca ti o n " ,
" f or e i g nP ro pe r t y ": null ,
" i sL oc at io n " : true ,
" g eo L oc at io n Co nf ig ": [
{
" key ": " g e o Lo ca ti on " ,
" t yp e O f Se le ct i o n ": " st ri ng " ,
" i nv er tC o o r d s ": true ,
" s i n g le Fi e l d Lo ca t io n ": "" ,
" d el im it er ": "," ,
" t yp e G e ol oc at i o n ": " Po int "
}]
}] ,
" d a ta Co n te n t F or N GS IL D Co nv e rs i o n ": [
{
" g eo Lo ca ti on ": " -5 .15 9 42 1 90 ,
-37 .36 057 650 " ,
" c re D at e " : "19 96 -11 -12 " ,
" s ch oo lC it y " : " N ata l " ,
" s ch oo lN am e " : " A th en eu " ,
" s ch oo lS iz e " : " Bi g "
},
...
]
}
Listing 3: Example of payload to convert data to NGSI-LD
through the Context approach.
tence services provided by Smart Geo Layers as mid-
dleware infrastructure.
The main validation use case was carried out at the
Public Prosecution Service of Rio Grande do Norte
(MPRN), Brazil. Aquedücte was used in the process
of extracting, filtering, and importing data using a
Web service provided by MPRN. Data consist of ge-
ographical coordinates in GeoJSON format and the
Basic Education Development Index (IDEB) of each
municipality of the state of Rio Grande do Norte.
Once already imported into Smart Geo Layers, these
data can be queried through a Web application built
atop the middleware to display them at a map with
each municipality and its corresponding IDEB.
The steps for importing data into Smart Geo Layers
(see Figures 3 and 4) are:
1. Select the import setup type: Context or Common
(default) approach
2. Select the type of data extraction of import setup:
from a file or from external API/Web service
3. Select the domain layer which each imported en-
tity will belong to
4. Setup for request to an external Web service and
data set loading
5. Select data from source to be handled
6. Filter fields for importing
7. Select fields for GeoJSON conversion (optional);
8. Convert to NGSI-LD protocol and visualize the
converted data
9. Import data into middleware by using the persis-
tence service
Figure 5 presents a Web application with the plot of
each city in the state of Rio Grande do Norte with
its respective educational indices as the final result
of the importation process using Aquedücte. The use
of Aquedücte was essential for using data/informa-
tion from external sources by Smart Geo Layers. As
it converted the imported data to the NGSI-LD pro-
tocol, they are available to applications that use any
NGSI-LD-based middleware.
5 RELATED WORK
We conducted a systematic literature review on the
extraction, integration, and data import within the
scope of smart cities. Major publication electronic
databases such as IEEEXplore, ACM Digital Library,
Scopus, ScienceDirect.com, and Web of Knowledge
were used to automatically retrieve studies. The fol-
lowing search string was used:
("data integration" OR "data extraction"
OR "data importation")
AND ("smart city" OR "smart cities")
The search process returned 47 studies. Most of the
selected works address data integration, extraction or
import aiming to meet a specific domain in smart
cities, such as transport, environment, and safety.
Three of them have drawn attention due to some sim-
ilarities with the proposal of Aquedücte.
(Fortini and Davis, 2018) address the context of
urban transportation. The extraction/collection of ur-
ban data provided by heterogeneous sources allows
the visualization of indicators and geometry of a given
city. For example, it allows comparing and evaluating
both public and private transport efficiency indicators.
A Linked Data-based Service for Integrating Heterogeneous Data Sources in Smart Cities
209

Figure 3: Aquedücte UI Workflow (Part 1/2).
(Mehmood et al., 2019) propose a data lake approach,
which would be supplied by diverse sets of data from
four pilot cities based on Big Data technologies, e.g.,
Hadoop File System, Spark, and Apache Flume. From
this data lake, it would be possible to analyze and vi-
sualize its data, thus providing more efficient decision
making in smart cities.
SmartLand-LD (Piedra and Suárez, 2018) is a
Linked Data-based framework that provides a flexi-
ble distributed ecosystem for data collection, extrac-
tion, and publication. It proposes an infrastructure to
achieve interoperability and data integration from di-
ICEIS 2020 - 22nd International Conference on Enterprise Information Systems
210

Figure 4: Aquedücte UI Workflow (Part 2/2).
Figure 5: A Web Application That Uses NGSI Converted
Data from smart Geo Layers.
verse sources that may exist in a smart city through
the use of ontologies. This concept is applied through
the conversion of the extracted data into the RDF
format. The framework also provides data/resources
availability through a well defined API to end-users
and several other applications.
These works have some points in common with
Aqueducte in terms of addressing data integration of
different formats and sources. However, the main dif-
ferences are (i) the use of the NGSI-LD protocol to
ease the integration of heterogeneous data and (ii) the
provision of a friendly UI that makes data extraction,
filtering, and import process easier for any middle-
ware or platform that works with the NGSI-LD pro-
tocol, as shown in the use case of MPRN (see Sec-
tion 4).
6 FINAL REMARKS
This paper presented Aquedücte, a service for han-
dling heterogeneous data integration in smart city ap-
plications. Aquedücte supports data extraction, filter-
ing, and import for any middleware that works with
the NGSI-LD protocol, to which data from different
formats and sources can be converted. Aquedücte was
A Linked Data-based Service for Integrating Heterogeneous Data Sources in Smart Cities
211

initially validated in a real-world scenario in conjunc-
tion with an NGSI-LD-compliant middleware plat-
form that allows integrating data provided by hetero-
geneous sources, correlating them to geographic in-
formation, and aggregating visualizing, querying, an-
alyzing these data.
Ongoing work is currently focused on improve-
ments on Aquedücte to support loading of large files
to have data extracted, filtered, and converted to the
NSGI-LD format. In addition, Aquedücte will operate
together with a microservice to perform user-defined
relationships of data to be imported, following the
NGSI-LD specification. Such features will be possi-
ble due to the adoption of a distributed file system,
which will allow managing such a large volume of
data in a more effective way.
REFERENCES
Bizer, C., Heath, T., and Berners-Lee, T. (2009). Linked
Data - the story so far. International Journal on Se-
mantic Web and Information Systems, 5(3):1–22.
Consoli, S., Presutti, V., Recupero, D. R., Nuzzolese, A. G.,
Peroni, S., Mongiovi’, M., and Gangemi, A. (2017).
Producing Linked Data for smart cities: The case of
Catania. Big Data Research, 7:1–15.
d’Aquin, M., Davies, J., and Motta, E. (2015). Smart cities’
data: Challenges and opportunities for semantic tech-
nologies. IEEE Internet Computing, 19:66–70.
Fortini, P. M. and Davis, C. A. (2018). Analysis, integra-
tion and visualization of urban data from multiple het-
erogeneous sources. In Proceedings of the 1st ACM
SIGSPATIAL Workshop on Advances on Resilient and
Intelligent Cities, page 17–26, New York, NY, USA.
ACM.
Heath, T. and Bizer, C. (2011). Linked Data: Evolving the
Web into a global data space. Morgan & Claypool
Publishers.
Lanthaler, M. (2013). Creating 3rd Generation Web APIs
with Hydra. In Proceedings of the 22nd International
Conference on World Wide Web, pages 35–38, New
York, NY, USA. ACM.
Mehmood, H., Gilman, E., Cortes, M., Kostakos, P., Byrne,
A., Valta, K., Tekes, S., and Riekki, J. (2019). Imple-
menting Big Data lake for heterogeneous data sources.
pages 37–44, USA. IEEE.
Piedra, N. and Suárez, J. P. (2018). SmartLand-LD: A
Linked Data approach for integration of heteroge-
neous datasets to intelligent management of high bio-
diversity territories. In Mejia, J., Muñoz, M., Rocha,
Á., Quiñonez, Y., and Calvo-Manzano, J., editors,
Trends and Applications in Software Engineering, vol-
ume 688 of Advances in Intelligent Systems and Com-
puting, pages 207–218. Springer International Pub-
lishing AG, Switzerland.
Rocha, B., Cavalcante, E., Batista, T., and Silva, J. (2019).
A Linked Data-based semantic information model for
smart cities. In Proceedings of the IX Brazilian Sym-
posium on Computing Systems Engineering, USA.
IEEE.
Souza, A., Pereira, J., Batista, T., Cavalcante, E., Cacho,
N., Lopes, F., and Almeida, A. (2018). A geographic-
layered data middleware for smart cities. In Proceed-
ings of the 24th Brazilian Symposium on Multimedia
and the Web, pages 411–414, New York, NY, USA.
ACM.
ICEIS 2020 - 22nd International Conference on Enterprise Information Systems
212