Knowledge Provenance in
Semantic Wikis
Li Ding, Jie Bao, and Deborah McGuinness
Tetherless World Constellation,
Rensselaer Polytechnic Institute
Troy, NY USA
Presented at AGU 2008 Fall Meeting
15–19 December 2008, San Francisco, California
Emerging Web Semantics
Web 1.0
Collaborative Authoring
• Web 1.0: online text
• Web 2.0: collaborative
online text
• Web 3.0: collaborative
online knowledge
Web 3.0
Web 2.0
Structure and Semantics Preservation
Motivation Question
What has been done to my homepage?
• Existing approaches
– Internet Archive: “when”
– Wiki history: “when”, “who”
• In Semantic Wiki
– Represent changes
– Query changes
Semantic History
• Automatically convert Wiki History to
Knowledge Provenance interlingua – PML
(Proof Markup Language). This links
semantic web data to information about
where it came from.
• We provided triple level granularity for
knowledge provenance in Semantic Media
Wiki
• We support semantic queries over
knowledge provenance as well as content
Encoding Wiki History Using PML
hasConclusion
Version 2 of Page A
NodeSet A2
User:Joe
hasInferenceEngine
isConsequentOf
hasInferenceRule
InferenceStep X
ex:Revision
hasAntecedent
hasCreationDatatime
17:17, October 22, 2008
hasConclusion
Version 1 of Page A
NodeSet A1
Knowledge Provenance
Interlingua
• PML1 was introduced in 2002
– It has been used in multiple contexts ranging from
explaining theorem provers to text analytics to machine
learning.
– It was specified as a single ontology
• PML2 improves PML1 by
– Adopting a modular design: splitting the original ontology
into three pieces: provenance, justification, and trust
• This improves reusability, particularly for
applications that only need certain explanation
aspects, such as provenance or trust.
– Enhancing explanation vocabulary and structure
• Adding new concepts, e.g. information
• Refining explanation structure
6
PML Provenance Ontology
• Scope: annotating
provenance metadata
• Highlights
– Information
– Source Hierarchy
– Source Usage
7
Initial Results
•(cur) (last) 17:17, October 22, 2008 Li (Talk | contribs | block) (420 bytes) (rollback | undo)
•(cur) (last) 20:42, August 24, 2008 Baojie (Talk | contribs | block) (345 bytes) (undo)
•(cur) (last) 23:44, August 15, 2008 Baojie (Talk | contribs | block) (331 bytes) (undo)
Ongoing Work
SMW History and Temporal RDF
Wiki Page “DLM”
Wiki Page “Triple 1”
Wiki Page “interval t1”
2008-11-12
begin-datetime
tw:DLM
rdf:subject
interval
Interval t1
end-datetime
foaf:firstName
rdf:predicate
Triple 1
2008-11-14
interval
rdf:object
Deborah
• Referring RDF triple via Reification
• integrating temporal data within RDF
Claudio Gutiérrez, Carlos A. Hurtado, Alejandro A. Vaisman - Temporal RDF. In ESWC pp. 93-107, 2005
Ongoing Work
Semantic History Computation
PO1 PO2 PO3 (triple, r_add, r_del)
R1 X
(po1,r1,null)
R2
X
(po1,r1,t2) (po2,r2, null)
R3 X
X
(po1,r3,null)
R4
X
R5 X
R6
X
X
(po1,r3,t4) (po3,r4,null)
X
(po1,r5,null) (po2.r2,r5)
Change of an RDF wiki page
POi:Property and object,
Ri: for revision id.
Space complexity DELTA
O(m)<= DELTA <=O(m*n)
m: # of triples
n: # of revisions
(po1,r5,r6) (po2,r6,null )(po3,r4,r6)
R7 X
(po1,r7,null) (po2,r6,r7)
Depends on change rate
Triple/revision
Rn
bitmap
delta
Link table
Alternative is bitmap
Rev_id, => linked template_page_id
non rdf:type (smw_rels2 ) S_id, P_id, O_id, rev_id_add, rev_id_del
rdf:type (smw_inst2) S_id, O_id, rev_id_add, rev_id_del
Revision Table Rev_id, rev_page, author, timestamp
Potential applications
Computing Trust using History
Trust in fragments
detected by analyzing
revision history (who
edited which fragment)
The lighter color
corresponds to low
trust
Conclusion and Future Work
• We have exposed wiki history information
about Semantic Web data using the PML
Knowledge Provenance Interlingua
• PML-based wiki history improves linking
and access to Semantic Media Wiki data
• We adopted a Temporal RDF approach
which is compatible with Semantic Query
capabilities offered by Semantic Media
Wiki
• The next version of Semantic history will
implement Temporal RDF in SMW