Transcript
Semantic Technologies for Linked Open Data at the STLab
Aldo Gangemi*, Andrea Nuzzolese, Valentina Presutti*, Diego Reforgiato, Alberto Salvati*,
Eva Blomqvist, Enrico Daga*, Francesco Draicchio, Paolo Ciancarini°, Sergio Consoli, Silvio Peroni°, Daria Spampinato*
CNR Semantic Technology Lab, ISTC-CNR, Rome/Catania name.surname@istc.cnr.it ; *name.surname@cnr.it ; °surname@cs.unibo.it
http://stlab.istc.cnr.it http://wit.istc.cnr.it/stlab-tools
http://data.cnr.it
People• STLab@ISTC
Aldo Gangemi Valentina Presutti Daria Spampinato Andrea Nuzzolese Diego Reforgiato Stefania Capotosti Sergio Consoli Alessio Iabichella
• STLab@SI Alberto Salvati
Gianluca Troiani
• STLab Associates
Paolo Ciancarini (UniBo)
Malvina Nissim
(UniBo)
Massi Ciaram
ita (Google)
Alfio Gliozzo (IBM)
Eva Blomqvist
(Un. Linköping)
Enrico Daga (Open University
)
Alessandro Adamou (Open Un.)
Francesco Draicchio (UniBo)
Francesco Antinucci (CNR)
• STLab (Semantic Technology Lab) è un laboratorio dell’ISTC (Istituto di Scienze e Tecnologie Cognitive) del CNR, con sedi a Roma e Catania, attivo anche a Bologna e Parigi
�2
Outline
•The Linked Open Data (LOD) of CNR !
•The Semantic Scout !
•Machine reading for the Semantic Web !
•Knowledge pattern discovery and usage
�3
A practical experience: data.cnr.it and the Semantic Scout
Joint work by STLab and the Information Systems unit of CNR Thanks to
Alberto Salvati, Enrico Daga, Gianluca Troiani, Andrea Pompili, Angelo Olivieri
Past collaboration with Claudio Baldassarre (UN-FAO) and Alfio Gliozzo (now IBM-Watson)
Linked Open Data in Public Administrations
�4
Objective and results• Objectives
• Publishing CNR data as LOD
• Matching the research demand to the research supply in the largest research institution (CNR) in Italy
!
• Results • data.cnr.it
• The CNR ontology network and data available as LOD
• Semantic interoperability between heterogeneous data sources
• The Semantic Scout - http://bit.ly/semanticscout • Expert finding based on competence
• Monitoring funding and evolution of different research areas and units
• Browsing and reporting capabilities
�5
�7
Methods for data conversion, extraction, inference, integration, linking, publishing, and searching
Semantic scoutExporting exploration results
�8 http://bit.ly/semanticscout
Machine reading for the Semantic Web
Apache Stanbol• A set of reusable components for semantic content
management • To extend traditional content management systems with semantic services
accessible as HTTP REST services
• Stanbol is the main software result of the EU IP IKS !
• Our contribution: the Knowledge Representation and Reasoning layer of Stanbol • Services used to define and manipulate semantic data models in CMS, i.e.,
Ontology Network Manager component
• Services able to retrieve additional semantic information about content, i.e., Reaoners and Rules components
�10
Stanbol in a nutshell
�11
NER and linking to LOD datasets
• The Black Hand might not have decided to barbarously assassinate Franz Ferdinand after he arrived in Sarajevo on June 28th, 1914
qualities
modality
negation
type induction WSD taxonomy induction
semantic roles
NER
events
dates
The <span xmlns:dbo="http://dbpedia.org/ontology/" xmlns:dbr="http://dbpedia.org/resource/" about="dbr:Black_Hand_(Serbia)" typeof=”dbo:Agent">Black Hand</span> might not have decided to barbarously assassinate <span xmlns:schemaorg="http://schema.org/" xmlns:dbr="http://dbpedia.org/resource/" about="dbr:Archduke_Franz_Ferdinand_of_Austria" typeof=”schemaorg:Person”>Franz Ferdinand</span> after he arrived in <span xmlns:schemaorg="http://schema.org/" xmlns:dbr="http://dbpedia.org/resource/" about="dbr:Sarajevo” typeof=”schemaorg:City”>Sarajevo</span> on June 28th, 1914
sample RDFa annotation
FRED http://wit.istc.cnr.it/stlab-tools/fred/
co-reference tense
�12�12
RESTful
Tìpalo
• Motivation
• It is difficult to automatically generate enterprise taxonomies from data available as plain documents
!
• Objective
• To enable automatic generate taxonomies by exploiting the richness of natural language text
�13
Typing DBpedia entities with Tìpalo
Typing
NER
Taxonomy induction
WSD Alignment to WordNet supersenses
Alignment to Dolce
http://wit.istc.cnr.it/stlab-tools/tipalo/ �14
“Pakito is the alias of french electronic dance music artist Julien Ranouil” (cf. wikipedia.org)
RESTful
Sentilo
• Sentilo is a new method of Sentic Computing
• i.e., Semantic Sentiment Analysis, which is a new research area
!
• Motivations
• Sentiment Analysis does not take into account semantic features when computing opinion scores
• Semantics can give a lot of information for Sentiment Analysis methods
!
• Objectives
• To provide Sentiment Analysis methods with Semantic information
• To identify more easily and also using semantic information the opinion
�15
Sentilo
�16
Topic
Sub topics
Opinion holder
Opinions
Sentiment scores
“Robert is happy because Silvio Berlusconi finally was
condemned by judges”
http://wit.istc.cnr.it/stlab-tools/sentilo /
RESTful
Knowledge pattern discovery
Bottom-up: schema extraction
�18
Encyclopedic Knowledge Patterns
• 184 Encyclopaedic Knowledge Patterns (EKPs) were discovered by identifying invariances in the structure of Wikipedia page links !
• EKPs are represented as OWL2 ontologies !
• They capture concepts that are typically used by Wikipedia users for describing things of a certain type
�19
An EKP for OfficeHolder
http://ontologydesignpatterns.org/ekp/�20
An EKP for OfficeHolder
Formal represenation
http://ontologydesignpatterns.org/ekp/�20
An EKP for OfficeHolder
Access to data
http://ontologydesignpatterns.org/ekp/�20
An EKP for OfficeHolder
Textual grounding
From wikipedia.org
http://ontologydesignpatterns.org/ekp/�20
Aemoo
• Aemoo exploits EKPs for • Entity summarisation and Exploratory search
• Distinguishing between core and peculiar knowledge
!
• The data sources are Wikipedia, DBpedia, Twitter, and GoogleNews !
• Aemoo is a KP-aware application • Benefits from KPs for addressing knowledge interaction tasks
• Uses KPs as the basic unit of mean for representing, exchanging, as well as reasoning with knolwedge
�21
Conclusions
�23
• We have provided a practical overview about how to build Linked Open Data !
• We have provided case studies and scenarios for exploiting Linked Data !
• We have shown Linked Data-compliant algorithms and tools
�24
Thank you!
top related