A Semi-automatic Ontology Acquisition Method for the Semantic Web Man Li, Xiaoyong Du, Shan Wang Renmin University of China, Beijing WAIM May 2012 SNU IDB Lab. Hye Chan, Bae
Outline Introduction SOAM Case Study Conclusion Discussion 2
Introduction The Semantic Web aims to add – Semantics – Better structure to the information 3
Introduction Success of Semantic Web depends on – The proliferation of ontologies – Pay more attention to the construction of ontologies 4 How do I construct the ontology?
Introduction Manual development of ontologies still remains a tedious and cumbersome task 5
Introduction A large amount of data about various domains are organized and stored in relational database 6
Introduction SOAM – Semi-automatic Ontology Acquisition Method – Based on data in relational database – Balance the cooperation between user contributions and machine learning Acquire ontology directly by using a group of rules Refine ontology according to lexical knowledge repositories (semi-automatically) 7
SOAM overview Step4: Acquire ontological instances based on refined ontological structure Step3: Refine the obtained ontological structure Step2: Acquire ontological structure according to the database schema information Step1: Capture the information about relational database schema 8
SOAM overview 9
Acquiring Ontological Structure Prior assumption – Relational schema is at least in 3NF We have 11 rules for acquiring ontological structure!! 10
Acquiring Ontological Structure Rule 1 R1 A1 A2 A3 11 R2 A1 A4 R3 A1 A5 A6 RiRi A1 A2 A3 A4 A5 A6 Class C i Equivalence
Acquiring Ontological Structure Rule 2 RiRi A1 A2 A3 12 RiRi A1 A2 A3 A4 RjRj A3 A5 A6 Class C i
Acquiring Ontological Structure Rule 2 13 RiRi A1 A2 R2 A2 A5 Class C i R1 A1 A3 A4
Acquiring Ontological Structure Rule 3 14 RiRi A1 A2 A3 RjRj A4 A5 Class C i Class C j A3 Inclusion dependency
Acquiring Ontological Structure Rule 4 15 RiRi A1 A2 A3 A4 RjRj A2 A3 A5 Class C i Class C j is-part-of has-part-of
Acquiring Ontological Structure Rule 5 16 RkRk A1 A2 RjRj A5 RiRi A1 A3 A4 Class C i Class C j
Acquiring Ontological Structure Rule 6 17 RlRl A1 A2 A3 RjRj A2 A6 RiRi A1 A4 A5 Class C i Class C j RkRk A3 A7 Class C k
Acquiring Ontological Structure Rule 7 18 RiRi A1 A2 A3 Class C i String Number Datatype property A1 A2 A3
Acquiring Ontological Structure Rule 8 19 RiRi A1 A2 A3 RjRj A1 A4 A5 Inclusion dependency Class C i subclass-of
Acquiring Ontological Structure Rule 1 (ref.) 20 RiRi A1 A2 A3 RjRj A1 A4 A5 Equivalence Class C j RiRi A1 A2 A3 A4 A5
Acquiring Ontological Structure Rule 9, 10, RiRi A1 A2 A3 Class C i A1 minCardinality=1 maxCardinality=1 NOT NULL : minCardinality = 1 UNIQUE : maxCardinality = 1
Refining Ontological Structures The obtained ontological structure is coarse Refining obtained ontology according to machine-readable – dictionaries – thesauri 22
Refinement algorithm The basic idea 1.A user wants to refine a concept in the ontology 2.The algorithm can help him find some similar lexical entries 3.The user can refine the concept according to the information 23 Concepts k most similar lexical entries
Similarity measures Lexical similarity – Edit distance method is used (LSim) Similarity in conceptual level – Considers the similarity about Super-concepts (SupSim) Sub-concepts (SubSim) 24
Case Study 25
Conclusion Gives a semi-automatic ontology acquisition method – Based on data in relational database Future work – Apply our approach in other domains – Do some researched on acquiring ontology from other resources Natural language text XML And so on 26
Discussion Strong point – More practical rules for real data in relational database? – Refinement using lexical repositories Weak point – No example Hard to understand the rules fully – Need to understand more about ontology languages OWL 27
Thank you!!! 28