Semantic Web of Things (SWoT) applications focus on providing a wide-scale interoperability that allows the sharing of IoT devices across domains and the reusing of available knowledge on the web. However, the application development is difficult because developers have to do various tasks such as designing an application, annotating IoT data, interpreting data, and combining application domains. To address the above challenges, this paper demonstrates SWoTSuite, a toolkit for prototyping SWoT applications. It hides the use of semantic web technologies as much as possible to avoid the burden of designing SWoT applications that involves designing ontologies, annotating sensor data, and using reasoning mechanisms to enrich data. Taking inspiration from sharing and reuse approaches, SWoTSuite reuses data and vocabularies. It leverages existing technologies to build applications. We take a hello world naturopathy application as an example and demonstrate an application development process using SWoTSuite. The demo video is available at URL http://tinyurl.com/zs9flrt.
In recent years, we have been witnessing a growing number of applications
exploiting sensors are becoming increasingly popular. However, existing applications
are largely specific to a domain and they lack methods to create innovative
cross-domain applications. Some examples of cross-domain SWoT applications [
The aim of our work is to reduce the development effort for prototyping SWoT
applications [
SWoTSuite architecture
In the following, we describe an architecture of SWoTSuite and its components.
Semantic annotator. It interacts with IoT devices that return data in
heterogeneous formats. It converts sensors metadata in an unified description to provide
further reasoning to overcome heterogeneity issues, similar to approach adopted
by Semantic Sensor Observation Service (SemSOS) [
Persistent storage. It stores the M3 ontologies, M3 datasets [
Knowledge manager. It is responsible for indexing, designing, reusing, and combining domain-specific knowledge that further update knowledge in the persistent storage. We have been building datasets to reference, classify, and reuse domain-specific knowledge that we call Linked Open Vocabularies for the Internet of Things (LOV4IoT)6. The LOVIoT provides domain ontologies, datasets, and rules that could be reused to design cross-domain SWoT applications. Reasoning engine. It infers high-level knowledge using a Jena inference engine and M3 rules. The M3 rules are a set of rules that are extracted from LOV4IoT and redesigned to be compliant with the M3 taxonomy.
Query engine. It executes queries and provides smart suggestions. The query engine has been implemented using ARQ, a SPARQL processor for Jena. The query engine loads the M3 ontologies, M3 datasets, knowledge deduced from the reasoning engine, and executes SPARQL queries in order to provide suggestions. 5 http://sensormeasurement.appspot.com/ 6 http://sensormeasurement.appspot.com/?p=ontologies
Query Engine
Suggestions on browser
Web Services SPARQL Queries Ontologies
Semantic IoT Data +
Inferred Knowledge
Rule-based Reasoning Engine Semantic Annotator
IoT Data
LOV4IoT dataset Knowledge Manager
TripleStore Semantic IoT data
Ontologies DataSets SWoTSuite: A Toolkit for Prototyping Cross-domain SWoT Applications
SPARQL Query Editor
Persistence Storage
SPARQL Queries
Rules For demonstration purposes, we take an application that combines data produced by a body thermometer with other domains such as healthcare and food knowledge bases. It suggests some preventive measures, such as home remedies when a fever is detected (e.g., drink orange or lemon juice because they contain Vitamin C) or consulting a doctor immediately in case of severity, that can be acted upon by patients. Application development using SWoTSuite consists of the following steps. The steps in even more detail can be found at URL7 and the recorded video is available at URL8.
Step 1: Generating an IoT application template. It selects a pre-defined template from the persistent storage. The template consists of ontologies, datasets, rules, and SPARQL queries that are required to build a SWoT application. To select a template, developers provide a set of sensors (e.g., body thermometer) and the related application domain (e.g., healthcare) information. SWoTSuite queries the persistent storage with this information and returns templates to select. The user interface is available at URL9.
Step 2: Semantically annotating data. Developers provide SenML data to
the semantic annotator to get the RDF/XML data. We provide two options
7 http://sensormeasurement.appspot.com/?p=end_to_end_scenario
8 http://tinyurl.com/zs9flrt
9 http://sensormeasurement.appspot.com/?p=m3api
for interacting with the semantic annotator10. Option 1 allows the developers
to enter a URL of data source; in option 2, the developers enter SenML data
manually and press the converter button to get annotated data. Currently, we
have been working on different ways of annotating sensor data. One of ways is
tools like Kino [
Step 4: Executing the query engine. It executes SPARQL queries and provides suggestions. We guide developers to use ARQ, a SPARQL query processor for Jena. The query engine loads the M3 ontologies, datasets, annotated data (Step 1), derived suggestions (Step 3), and SPARQL queries. Step 5: Displaying results. It receives results from the query engine (Step 4) and displays derived knowledge and suggestions on a Web browser13.
We have evaluated our framework on three aspects: the performance of the
semantic annotator given a size of data [1, p. 85], the performance of the reasoner
given a number of rules [1, p. 86], and the performance of different tasks according
to a size of data [1, p. 87]. We are going to evaluate the usability of SWoTSuite
with real users through a Google form at our ISWC 2016 tutorial [