Se ha denunciado esta presentación.
Utilizamos tu perfil de LinkedIn y tus datos de actividad para personalizar los anuncios y mostrarte publicidad más relevante. Puedes cambiar tus preferencias de publicidad en cualquier momento.
Technische Universität Chemnitz  Prof. Dr.-Ing. Martin Gaedke & Team 23.10.2018
Concept Extraction from theWeb
ofThings K...
Introduction
One of the six technologies that will most influence the world by 2025
IoT market will grow to $7.1 trillion
>300 IoT plat...
Challenge: missing interoperability
#40% of the potential benefits of
IoT cannot be obtained without
interoperability
McKi...
“Enablingdifferentagents,services, andapplicationsto
exchange information,dataandknowledge in a
meaningfulway, on andoffth...
Automatically
built
my machines
Automatically built
by machines
Google’s Knowledge
Graph
You are using the SemanticWeb Eve...
Ongoing
Extension to
schema.org
iot.schema.org
How to reuse the domain knowledge already encoded in
IoT ontologies?
Research Challenge
10
cc
RelatedWork
Ontology Analysis
semantic interoperabilitypractices
(EuropeanResearchCluster,2015)
health,transportation and logistic
(Ga...
Approach Input Search terms
Tool
Availability
IoT ontology
Support
Output
OntoKhoj,
Sindice
keyword class, subclass, domai...
13
Approach
Identifying Most Popular Concepts
STEP 1: Ontology Selection
• Selection of 14 IoT ontologies
̶ Standardized
̶ The most cited when investigating the literat...
STEP 2: Preprocessing
Convert all
ontologies to
ttl format
ASCII
Conversion
Case
Conversion
Store
ontologies in
Virtuoso
STEP 3: Vocabulary Extraction
STEP 4: Term Extraction
Split vocabulary
on camel-case
or snake-case
Remove
punctuation
marks
Remove
numbers
Stop Word
Rem...
STEP 5: Frequency Calculation
Given a set of terms 𝑡 ∈ 𝑇, the frequency of the term in
the set of ontologies 𝑂 is calculat...
STEP 6: Word2Vec
STEP 7: k-Means Clustering
STEP 8: Results Aggregation
Using the term frequencies and term clusters as input this
step calculates the total number of...
STEP 9: Concept Assignment
Given the term clusters the ontology experts
manually assigned names to each cluster based on
t...
24
Results
Identification of Popular Concepts
Extent of re-use among ontologies
27
Conclusion
Concept Extraction from IoT Ontologies
Existing knowledge is constantly redesigned in different
communities
Extent of reus...
VSR
Technische Universität Chemnitz  Prof. Dr.-Ing. Martin Gaedke & Team 23.10.2018
ThankYou!
mahda.noura@informatik.tu-c...
Próxima SlideShare
Cargando en…5
×

Concept extraction from the web of things (3)

Mahda Noura, Amelie Gyrard, Sebastian Heil and Martin Gaedke. Concept extraction from the web of things knowledge bases. International Conference WWW/Internet. 21-23 October 2018, Budapest, Hungary,

Paper: http://knoesis.org/node/2913
Semantic web technologies are a major driver for semantic interoperability in IoT-generated data by using shared vocabularies in an ontology-driven approach. While there is a growing interest in standardization of ontologies for IoT, there is still a lack of common agreement for a specific IoT ontology. Numerous concepts and relations have been designed within existing ontologies to handle different features of IoT data. However, there are many redundant and overlapping concepts designed within existing standardizations and groups. We found that new ontologies constantly redesign the same concepts in IoT. Therefore, it is a challenge to reuse and unify these different IoT ontologies with redundant concepts. In this paper, we investigate what are the most used terms within IoT ontologies? We identify the fourteen most popular ontologies within generic IoT and WoT domain. Analysis of popular concepts among these ontologies allows to automatically rank the knowledge. This work will enable guiding ontology engineers to re-use and unify existing ontologies, a required step to achieve semantic interoperability. Moreover, this work could contribute towards building iot.schema.org.

  • Sé el primero en comentar

Concept extraction from the web of things (3)

  1. 1. Technische Universität Chemnitz  Prof. Dr.-Ing. Martin Gaedke & Team 23.10.2018 Concept Extraction from theWeb ofThings Knowledge Bases Mahda Noura*,Amelie Gyrard**, Sebastian Heil*, Martin Gaedke* *Technische Universität Chemnitz,Chemnitz,Germany ** Knoesis,Wright State University, Ohio, USA
  2. 2. Introduction
  3. 3. One of the six technologies that will most influence the world by 2025 IoT market will grow to $7.1 trillion >300 IoT platforms in the market IDC UNIFY IoT NIC Internet ofThings 3
  4. 4. Challenge: missing interoperability #40% of the potential benefits of IoT cannot be obtained without interoperability McKinsey 4
  5. 5. “Enablingdifferentagents,services, andapplicationsto exchange information,dataandknowledge in a meaningfulway, on andoffthe Web“ (W3C) 5 Semantic Interoperability
  6. 6. Automatically built my machines Automatically built by machines Google’s Knowledge Graph You are using the SemanticWeb Every Day!
  7. 7. Ongoing Extension to schema.org iot.schema.org
  8. 8. How to reuse the domain knowledge already encoded in IoT ontologies? Research Challenge
  9. 9. 10 cc RelatedWork
  10. 10. Ontology Analysis semantic interoperabilitypractices (EuropeanResearchCluster,2015) health,transportation and logistic (Ganzhaetal.2017) ontology validationtools (Gyrard etal.2018) 11
  11. 11. Approach Input Search terms Tool Availability IoT ontology Support Output OntoKhoj, Sindice keyword class, subclass, domain/range ✗ ✗ ranked ontologies Swoogle, Watson keyword class, property, label, comment, literal ✓ ✗ ranked ontologies AKTiveRank keyword terms ✗ ✗ ranked ontologies Falcons keyword class, property, labels ✗ ✗ ranked ontologies OntoSelect, OntoSearch2 keyword class, label, property, ontology-title ✗ ✗ ranked ontologies Popular Concepts in IoT ontologies? not considered
  12. 12. 13 Approach
  13. 13. Identifying Most Popular Concepts
  14. 14. STEP 1: Ontology Selection • Selection of 14 IoT ontologies ̶ Standardized ̶ The most cited when investigating the literature survey ̶ Availability of the ontology, code online ̶ Ability to work with a set of validation tools (e.g., PerfectO)
  15. 15. STEP 2: Preprocessing Convert all ontologies to ttl format ASCII Conversion Case Conversion Store ontologies in Virtuoso
  16. 16. STEP 3: Vocabulary Extraction
  17. 17. STEP 4: Term Extraction Split vocabulary on camel-case or snake-case Remove punctuation marks Remove numbers Stop Word Removal
  18. 18. STEP 5: Frequency Calculation Given a set of terms 𝑡 ∈ 𝑇, the frequency of the term in the set of ontologies 𝑂 is calculated 𝑓𝑂(𝑡) as it represents the total number of occurrences of each unique term in all ontologies
  19. 19. STEP 6: Word2Vec
  20. 20. STEP 7: k-Means Clustering
  21. 21. STEP 8: Results Aggregation Using the term frequencies and term clusters as input this step calculates the total number of occurrences of the terms in different ontologies as well as the ontologies using this term per cluster
  22. 22. STEP 9: Concept Assignment Given the term clusters the ontology experts manually assigned names to each cluster based on the semantic meaning of the cluster.
  23. 23. 24 Results
  24. 24. Identification of Popular Concepts
  25. 25. Extent of re-use among ontologies
  26. 26. 27 Conclusion
  27. 27. Concept Extraction from IoT Ontologies Existing knowledge is constantly redesigned in different communities Extent of reuse among ontologies is low Future Work Create training dataset for IoT Apply the presented methodology to other IoT application domains
  28. 28. VSR Technische Universität Chemnitz  Prof. Dr.-Ing. Martin Gaedke & Team 23.10.2018 ThankYou! mahda.noura@informatik.tu-chemnitz.de VSR.Informatik.TU-Chemnitz.de @mahdanoura /mahda noura /mahdanoura

×