312
M. Noura et al.
Ontology Selection: 16 smart car ontologies are collected from LOV4IoT for evaluation purposes (Table 1). The most important ontologies in each domain have been
selected according to the following criteria:
• Citations of the scientific publications describing the ontology (e.g., the SSN ontology v1 [62] has more than 1000 citations): higher is the number, better the ontology
might be. However, this criteria cannot be applied to recent publications.
• Journal impact factor and conference ranking: higher the ranking is, better would be
the ontologies. Within the references section, the ranking is added for publications
cited and classified within Tables 1 and 2.
• Recent publications increase the chance to have the authors maintaining the ontology and integrating previous ontologies.
• Ontologies disseminated in standardizations (e.g., W3C Web of Things ontology,
16 W3C SSN/SOSA ontology [63], ETSI M2M SAREF ontology [64]) can be
considered as more reliable.
• Industrial partners involved, the project is considered more impactful, and the
implementation is more reliable.
• Domain experts involved (not computer scientists) since they share their human
expertise.
• Ontology code that can be downloaded, because, in science, the experiments should
be replicable, following the FAIR (Findability, Accessibility, Interoperability, and
Reuse) principles.
Ground Truth Dataset Design: Domain experts can participate in the questionnaire
to design the ground truth (a similar questionnaire for smart cities, weather, and
smart home is available online,
17 see [61]) for detailed information). Experts were
either involved in developing smart car ontologies or are an open audience having
the domain expertise to describe each ontology using three keywords. The participants’ level of expertise in the automotive domain and knowledge engineering, is
asked in a Likert scale of five levels, from ‘totally disagree’ to ‘totally agree’. The
experts are given the list of ontologies (through a series of figures from the ontology
classes in Protege) in different domains to select the top three keywords that best
describes that ontology in relation to the keywords that were obtained from the main
concepts in the generated clusters. Domain experts chose keywords among a total
number of keywords in our evaluation form.
6 Conclusion and Future Work
The Systematic Literature Survey (SLS) in any research topics is a time-consuming
approach. Finding knowledge returned by Google results still require a huge work
on learning, classification and summarizing. To ease this time-consuming task, we
16 https://www.w3.org/TR/wot-thing-description/.
17 https://bildungsportal.sachsen.de/survey/limesurvey/index.php/716626/lang-en.
M. Noura et al.
Ontology Selection: 16 smart car ontologies are collected from LOV4IoT for evaluation purposes (Table 1). The most important ontologies in each domain have been
selected according to the following criteria:
• Citations of the scientific publications describing the ontology (e.g., the SSN ontology v1 [62] has more than 1000 citations): higher is the number, better the ontology
might be. However, this criteria cannot be applied to recent publications.
• Journal impact factor and conference ranking: higher the ranking is, better would be
the ontologies. Within the references section, the ranking is added for publications
cited and classified within Tables 1 and 2.
• Recent publications increase the chance to have the authors maintaining the ontology and integrating previous ontologies.
• Ontologies disseminated in standardizations (e.g., W3C Web of Things ontology,
16 W3C SSN/SOSA ontology [63], ETSI M2M SAREF ontology [64]) can be
considered as more reliable.
• Industrial partners involved, the project is considered more impactful, and the
implementation is more reliable.
• Domain experts involved (not computer scientists) since they share their human
expertise.
• Ontology code that can be downloaded, because, in science, the experiments should
be replicable, following the FAIR (Findability, Accessibility, Interoperability, and
Reuse) principles.
Ground Truth Dataset Design: Domain experts can participate in the questionnaire
to design the ground truth (a similar questionnaire for smart cities, weather, and
smart home is available online,
17 see [61]) for detailed information). Experts were
either involved in developing smart car ontologies or are an open audience having
the domain expertise to describe each ontology using three keywords. The participants’ level of expertise in the automotive domain and knowledge engineering, is
asked in a Likert scale of five levels, from ‘totally disagree’ to ‘totally agree’. The
experts are given the list of ontologies (through a series of figures from the ontology
classes in Protege) in different domains to select the top three keywords that best
describes that ontology in relation to the keywords that were obtained from the main
concepts in the generated clusters. Domain experts chose keywords among a total
number of keywords in our evaluation form.
6 Conclusion and Future Work
The Systematic Literature Survey (SLS) in any research topics is a time-consuming
approach. Finding knowledge returned by Google results still require a huge work
on learning, classification and summarizing. To ease this time-consuming task, we
16 https://www.w3.org/TR/wot-thing-description/.
17 https://bildungsportal.sachsen.de/survey/limesurvey/index.php/716626/lang-en.