=Paper=
{{Paper
|id=Vol-2400/paper-18
|storemode=property
|title=Using a Smart City Ontology to Support Personalised Exploration of Urban Data (Discussion Paper)
|pdfUrl=https://ceur-ws.org/Vol-2400/paper-18.pdf
|volume=Vol-2400
|authors=Devis Bianchini,Valeria De Antonellis,Massimiliano Garda,Michele Melchiori
|dblpUrl=https://dblp.org/rec/conf/sebd/BianchiniAGM19
}}
==Using a Smart City Ontology to Support Personalised Exploration of Urban Data (Discussion Paper)==
https://ceur-ws.org/Vol-2400/paper-18.pdf
Using a Smart City Ontology to support
Personalised Exploration of Urban Data
(Discussion Paper)
Devis Bianchini1 , Valeria De Antonellis1 , Massimiliano Garda1 , and Michele
Melchiori1
University of Brescia, Dept. of Information Engineering
Via Branze 38, 25123 - Brescia (Italy)
devis.bianchini@unibs.it, valeria.deantonellis@unibs.it,
m.garda001@unibs.it, michele.melchiori@unibs.it
Abstract. During the latest years, Smart City projects aimed at driv-
ing local governments towards the strong use of technologies to support
a higher quality of urban spaces and a better offering of public services.
From an information viewpoint, this means enabling different categories
of users, including citizens, Public Administration (PA), utility and en-
ergy providers, to access the large amounts of data in multiple, hetero-
geneous Smart City data sources, by adopting new tools and methods to
take decisions that might improve city daily life. Aggregation of urban
data according to multiple perspectives through the definition of proper
indicators enables urban data exploration at different granularity levels
for distinct categories of users. Furthermore, Semantic Web technolo-
gies may be used to enable interoperability and improve data access. In
this paper, we propose a Smart Living Ontology, to provide a semantic-
enriched representation of city indicators. On top of the ontology and
users’ characterisation, a Semantic Layer has been designed to enable
personalised access to urban data.
Keywords: urban data exploration, semantic web, smart city ontology
1 Introduction
Smart City projects aim at driving local governments towards the strong use of
technologies to support a higher quality of urban spaces and public services [1–3].
From an information viewpoint, different categories of users, including citizens,
Public Administration (PA), utility and energy providers, must explore the large
amounts of data from multiple, heterogeneous data sources, in order to take
decisions that might improve city daily life. In recent research, Semantic Web
Copyright c 2019 for the individual papers by the papers’ authors. Copying per-
mitted for private and academic purposes. This volume is published and copyrighted
by its editors. SEBD 2019, June 16-19, 2019, Castiglione della Pescaia, Italy.
technologies have been proposed to develop ontology-enabled applications, such
as Smart Urban Cockpits and dashboards [8, 10], where proper indicators have
been semantically described [6]. Indicators aggregate urban data according to
several perspectives and provide a comprehensive view over underlying data
without being overwhelmed by the data volume [7].
In this paper, we propose a semantics-enabled framework, which relies on the
so-called Smart Living Ontology, to provide a semantic representation of Smart
City indicators. The framework also includes the kinds of activities and users’
categories for which indicators have been designed to share relevant information.
A Semantic Layer, developed on top of the ontology, enables personalised ex-
ploration of urban data. The framework has been designed for decision makers,
who need to have a view on heterogeneous urban data at different aggregation
levels, ranging from energy consumption to garbage collection, pollution lev-
els, citizens’ safety. For example, the framework may allow building managers to
monitor electrical consumption of administered buildings, by exploiting the indi-
cators hierarchy in the ontology to distinguish electrical consumption according
to different perspectives (e.g., consumption in common spaces, consumption of
elevators), and to compare average values of consumption with other buildings at
district or city level. Furthermore, the framework may enable citizens to make
decisions about their activities by observing specific indicators (e.g., to avoid
sport activities when pollution levels overtake tolerance thresholds).
This work was performed in the context of the Brescia Smart Living (BSL)
Italian project1 , which promotes a holistic view of the city, where different types
of data are explored to provide new services to both citizens and PA. The frame-
work has been already presented in [5], where more details about the Smart
Living Ontology and the implementation have been described.
This paper is organised as follows: in Section 2 we highlight the cutting-edge
features of our approach compared to the literature; in Section 3, motivations
are presented; Section 4 presents the Smart Living Ontology; Section 5 describes
the Semantic Layer of the framework; in Section 6 we discuss implementation
details and preliminary validation; finally, Section 7 closes the paper.
2 Related Work
In literature, the adoption of ontologies in Smart City projects targets energy
management, where diagnostic models are built to discover energetic losses [11]
or to perform optimisation for cost saving [3]; facility discovery, to search for
city facilities and services [2]; events monitoring and management [1]; Ontology-
Based Data Access (please refer to [12] for a survey and related flagship research
projects), to cope with heterogeneous data sources inside the Smart City [4].
With respect to an OBDA perspective, our objective is to focus on data
exploration by exploiting a semantic characterisation of Smart City indicators,
also considering users’ category and activities for which indicators have been
1
http://www.bresciasmartliving.eu
designed. The semantic modelling we pursue also reinforces the characteristics
of the BSL project, that if compared to other Smart City projects [1–3] provides
a wider spectrum of urban data. Approaches focused on Ontology-Based Data
Warehouses (OBDW) store analytical data, indicators, requirements and their
semantics [10] or provide a semantic description of metrics used to compute indi-
cators [6], in order to enable meaningful comparison among different aggregated
measures. Differently from the aforementioned solutions, our approach is focused
on the exploitation of indicators hierarchies and dimensional modelling to guide
exploration of aggregated data for multiple categories of users, introducing a se-
mantic relationship between users’ profiles and indicators to foster personalised
urban data exploration.
3 Motivating scenario
As a motivating example, let’s consider John, the manager of several buildings
located in different districts of the Smart City. John monitors the electrical
consumption of the buildings, in order to implement energy saving policies (e.g.,
introducing LED lamps in common areas or planning renovation work to increase
the energy efficiency class of buildings). Challenging issues are related to the
capability of enabling John to fruitfully exploit available information. In this
paper, we address such issues as follows.
Semantic specification of city indicators. Indicators are commonly defined
to aggregate data according to several dimensions, for different categories
of users (e.g., consumption-based indicator of electrical energy use). In our
framework, the Smart Living Ontology is defined to provide semantic specifi-
cation of the indicators, that takes into account indicators scope, in terms of
spatial and temporal constraints, their hierarchical organisation and target
users.
Personalised data exploration. Given the variety of urban data that can be
explored, the selection of proper indicators is personalised taking into ac-
count users’ profiles, composed of user’s category, activities and preferential
indicators.
Indicators recommendation to support decision making. Indicators rec-
ommendation is provided in order to help users to take decisions in their
daily life. For example, John is provided with suggestions about indicators
on electrical consumption of the administered buildings. To this aim, indica-
tors scope and hierarchy, as well as filtering based on activities for which the
indicators have been designed, are exploited to better focus the exploration.
4 The Smart Living Ontology
Figure 1 reports the main concepts and relationships of the Smart Living Ontol-
ogy (SLO2 ). In order to face the heterogeneity and complexity of the Smart City
2
The TBox of the ontology can be found at https://tinyurl.com/onto-schema (a
free Web Protégé account is required)
Legend
qb: Constraint
Dimension SLO Imported
Property Concept Concept
Dimension time:
Instant ObjectProperty
time:hasEnd
User DatatypeProperty
time:hasBeginning Category Literal
time:Temporal time: rdfs:subClassOf
Entity Interval
Domain
isFor
boundTo
hasTimeGranularity
relatedTo
schema:
Place
hasSpatialCoverage
Indicator
influencedBy schema:
Activity Action
qb:Measure
schema:
Property hasBSLService
Administrative
Area
URL Environmental
hasSubIndicator Indicator
Leisure Building
schema: Energy … Administration
District
City Consumption
schema:contains … Pollution
Place
Indicator
schema:contains Indicator
Place
Electrical Energy Particulate
Apartment Building … Water Consumption
Workplace Consumption Concentration
Indicator … Indicator Indicator
Fig. 1. A portion of the Smart Living Ontology (SLO).
domain, we rely on some foundation ontologies to cover a set of required pivotal
concepts: (i) a geospatial mapping of the main structures of the city (e.g., build-
ings, streets, areas) and their topology; (ii) temporal entities; (iii) other high
level concepts, that have been specialised to define the hierarchy of indicators
and activities.
Indicators are specified as individuals of the Indicator concept or one of its
sub-concepts in the indicators hierarchy. An indicator is further relatedTo a set
of domain individuals (e.g., environment, safety, energy, mobility) to define the
indicator scope, and a set of constraints. As shown in Figure 1, in the SLO a
constraint can be either a dimension (time and space) or a user’s category (e.g.,
building manager). Specifically, an indicator can be boundTo a time interval
(e.g., values of electrical consumption available for the year 2017), may have
a time granularity (hasTimeGranularity relationship), may be defined at city,
street, district or more specific levels, such as buildings (hasSpatialCoverage
relationship). Finally, knowledge about an indicator can be useful to perform
specific activities, defined as individuals of the concept Activity or its sub-
concepts (influencedBy property). An indicator is linked to a web-based service
of the BSL Platform (hasBSLService property) to display the indicator values
on the Smart City Dashboard, as explained in Section 6.
BSL users are profiled according to their category, their activities, the types of
indicators explored by the user through the interactions with the framework. In
the next section we detail how personalised exploration of indicators is performed
based on the SLO, with the help of the motivating example.
5 Semantic Layer for Personalised Data Exploration
Personalised data exploration for each user can be achieved by effectively ex-
ploiting city indicators, properly selected according to the indicators domains,
constraints and profiles of users. Figure 2 reports the steps for semantics-enabled
data exploration and the I/O for each step.
Smart Living Ontology
user’s BSL Platform Web services
profile
request for
indicators
available
indicators
Semantics-enabled Data visualisation
Candidate Indicators
Personalised Data on the web-based
Selection
Exploration Dashboard
domain-driven
inputs indicators selection
outputs
activity-based
indicators refinement
filtering based on candidate refined candidate
user’s category indicators indicators
Fig. 2. The steps of semantics-enabled data exploration.
For example, let’s consider again the user John in the motivating example,
who is the manager of three buildings (namely Building 1, Building 2 and
Building 3) located in two districts of the city. Since John is usually interested
in monitoring buildings, during the registration to the BSL platform, he specifies
the activity Monitoring in his profile, jointly with his administered buildings,
associating them to the districts they are located in.
Candidate indicators selection. In order to have an insight on the status of the
buildings, for instance to evaluate whether replacing standard lamps with less
energy-demanding LED ones, John issues a request to the framework. To sup-
port John in the request formulation, without requiring a detailed knowledge of
ontology concepts and individuals, the framework enables him to specify a set
of keywords Kr = {energy, consumption}, processed according to techniques
aimed to match the keywords with ontology terms [9] (precisely, individuals of
Domain and Indicator concepts or sub-concepts). The platform processes the
request and returns, among the others, the indicator NormalizedElectrical-
EnergyConsumption (NEEC), which reports electrical consumption normalised
with the number of apartments in the building. The indicator is selected be-
cause it is compliant with the keywords given in the request (domain-driven
indicators selection), it is associated with the activity Monitoring in the ontol-
ogy (activity-based indicators refinement) and it is compliant with the building
manager category (filtering based on user’s category). User’s exploration history
(in terms of formerly inspected indicators) is traced and it is taken into ac-
count during indicators suggestion, as it can be exploited to assess the degree of
compliance between proposed indicators and user’s past exploration preferences.
Figure 3 reports the portion of the SLO containing the candidate indicators.
Semantics-enabled personalised data exploration. Starting from NEEC indicator
(Figure 3), John can further explore other indicators being guided by the se-
mantic relationships in the SLO. Exploration can be performed: (a) over the
indicators hierarchy and/or (b) over the indicators dimensions. In the former,
John selects the NEEC indicator and the framework suggests him more specific
indicators (following the hasSubIndicator relationship). Exploration over the
Downtown San Polino Monitoring
district district
schema:contains
schema:contains schema:contains
Place
Place Place
Legend
Energy
Building 1 Building 3 influencedBy
Concept Building 2
Individual hasSpatialCoverage
relatedTo URL
hasSpatialCoverage
hasBSLService
Literal
hasSpatialCoverage NormalizedElectricalEnergy Electrical
ObjectProperty Consumption (NEEC) rdfs:comment consumption at
building level
DatatypeProperty hasSubIndicator rdf:type
isFor
rdfs:subClassOf NEEC_Stairs hasSubIndicator hasSubIndicator Building
Energy
Consumption Manager
NEEC_Gardens Indicator
NEEC_Elevators rdf:type
rdf:type
rdf:type
Fig. 3. Example of candidate indicator and related properties.
indicators dimensions exploits both the knowledge on the spatial coverage of
indicators and the information stored in the user’s profile. Starting from indi-
cators previously selected for the John’s building, the containment relationship
that relates John’s buildings with districts is exploited. Therefore, John could
compare his buildings against others having similar characteristics or using dif-
ferent lighting solutions; this may stimulate John to consider the replacement of
energy consuming light bulbs with modern LED lamps in shared spaces.
6 Implementation and Preliminary Validation
Figure 4 shows the web-based architecture of the semantics-enabled data explo-
ration framework. The architecture is organised over three layers. Data on field,
collected from domain-specific platforms through IoT technologies, as well as
data from external sources (e.g., weather and pollution data), are loaded into
the BSL platform (BSL platform Layer). Data is transferred to the BSL platform
using RESTful services, SOAP-based services and MQTT Agents. Data is aggre-
gated into smart city indicators, which are semantically specified in the Semantic
Layer. The User Access Layer includes a web-based Smart City Dashboard to be
used by citizens, PA and other users to explore data and take decisions (see [5]
for more details). Using the web browser, users can register themselves and up-
date their profile. The framework is implemented in Java and deployed under the
Apache TomEE application server. The Smart Living Ontology is deployed in
OWL using the Stardog3 Triplestore. The Stardog module supports domain ex-
perts to maintain the ontology (concepts, relations and individuals), interacting
with the web-based administration console provided by the module.
Preliminary experiments on the proposed framework, aimed at demonstrat-
ing its effectiveness in supporting candidate indicators selection, have been con-
ducted considering a SLO composed of 57 concepts (among them, 30 indicators),
104 individuals, 207 object and datatype properties.
3
https://www.stardog.com/
RESTfulSer
vices Fi
eldI
oT(smar
tmeters,nextgenerat
iongasmet
ers
,
SOAPSer vi
ces hydr
oni
cval
ves,wear abledevi
ces,.
..
)
MQTTAgents andotherdat
as our
ces
Fig. 4. Web-based architecture of the semantics-enabled data exploration framework.
We considered two kinds of requests: (A) requests where the user specified
a set of keywords Kr in order to identify desired domains and indicators, and
the user’s profile does not contain any activity or preferential indicator; (B)
requests where the user presents a richer profile (containing category, activities
and preferential indicators), but specifies keywords in Kr , that only correspond
to individuals of the Domain concept. We compared our ontology-based approach
against a keyword-based search, where semantic disambiguation techniques have
been applied to Kr [9], but SLO semantic relationships have not been exploited.
Average precision and recall values for the keyword-based search are equal to 0.49
and 0.97 for type A (0.33 and 0.27 for type B, resp.), whereas for the ontology-
based search they are equal to 0.99 and 0.98 for type A (0.94 and 0.93 for type
B, resp.). Candidate indicators selection average execution time for type A is
about 2559 ms, whereas for type B is about 1325 ms. Since both the compared
approaches use keywords disambiguation techniques and the same keywords have
been used during tests, difference in average precision and recall is due to the
knowledge structure in the ontology. Usability tests are being performed to check
the capability of the framework in facilitating user’s access to urban data through
the suggestion of candidate indicators. To perform usability tests, we considered
metrics such as the number of exploration steps needed to obtain desired data,
number of fails, number of successful explorations. Currently, the framework is
being tested, with satisfaction, by a sample of users in two districts, a modern
one, where new generation smart meters have been installed, and a district in
city downtown, more densely populated and presenting older buildings. Usability
experiments are being carried on within the Brescia Smart Living project until
September 2019.
7 Conclusions
In this paper, we described a semantics-enabled framework composed of: (i)
a so-called Smart Living Ontology, apt to provide a flexible representation of
Smart City indicators; (ii) a Semantic Layer, to enable personalised exploration
of urban data for different categories of users. The framework has been already
presented in [5], where more details about the Smart Living Ontology and the
implementation have been described. Ontologies represent knowledge structure
that can be used to facilitate urban data exploration at different granularity
levels and according to different exploration perspectives. Future effort will be
devoted to extend the set of semantic relationships in the SLO as follows: (a)
further relationships between indicators will be identified (e.g., to assert that two
or more environmental indicators must be jointly monitored due to their harmful
impact on the ecosystem); (b) strategies to dispense useful recommendations for
promoting the users’ virtuous behaviours, providing advice for healthy activities
that should be practised by users.
References
1. The Res Novae project. http://resnovae-unical.eu.
2. The San Francisco Park project. http://sfpark.org.
3. The BESOS project. Building Energy decision Support systems fOr Smart cities.
http://besos-project.eu.
4. Bellini, P., Benigni, M., Billero, R., Nesi, P., Rauch, N.: Km4city ontology building
vs data harvesting and cleaning for smart-city services. Journal of Visual Languages
& Computing 25(6), 827 – 839 (2014)
5. Bianchini, D., De Antonellis, V., Garda, M., Melchiori, M.: Semantics-enabled
personalised urban data exploration. In: Proc. of the 19th International Conference
on Web Information Systems Engineering (WISE 2018). pp. 361–376 (2018)
6. Diamantini, C., Potena, D., Storti, E.: Comparison of City Performances Through
Statistical Linked Data Exploration. In: Proc. of the 2nd International Conference
on ICT Infrastructures and Services for Smart Cities (IISSC 2017). pp. 3–12 (2017)
7. Lanza, J., Sotres, P., Sánchez, L., Galache, J., Santana, J., Gutiérrez, V., Munoz,
L.: Managing large amount of data generated by a smart city internet of things de-
ployment. International Journal on Semantic Web and Information Systems 12(4),
22–42 (2016)
8. Psyllidis, A.: Ontology-Based Data Integration from Heterogeneous Urban Sys-
tems: A Knowledge Representation Framework for Smart Cities. In: Proc. of the
14th International Conference on Computers in Urban Planning and Urban Man-
agement (CUPUM 2014) (2015)
9. Royo, J.A., Mena, E., Bernad, J., Illarramendi, A.: Searching the web: From key-
words to semantic queries. In: Third Int. Conf. on Information Technology and
Applications (ICITA 2005). pp. 244–249 (2005)
10. Santos, H., Dantas, V., Furtado, V., Pinheiro, P., McGuinness, D.L.: From data to
city indicators: A knowledge graph for supporting automatic generation of dash-
boards. In: Proc. of the 14th European Semantic Web Conference (ESWC 2017).
Lecture Notes in Computer Science, vol. 10250, pp. 94–108 (2017)
11. Tomašević, N.M., Batić, M.v., Blanes, L.M., Keane, M.M., Vraneš, S.: Ontology-
based facility data model for energy management. Adv. Eng. Inform. 29(4), 971–
984 (2015)
12. Xiao, G., Calvanese, D., Kontchakov, R., Lembo, D., Poggi, A., Rosati, R., Za-
kharyaschev, M.: Ontology-based Data Access: A survey. In: Proc. of the 27th
International Joint Conference on Artificial Intelligence (IJCAI 2018). pp. 5511–
5519 (2018)