keynote for SMAP 2013

1. bridging formal semantics and social semantics on the web
010101
HELLO
Fabien Gandon, @fabien_gandon, http://fabien.info
Leader of the Wimmics research team (INRIA, CNRS, Univ. Nice)
W3C Advisory Committee Representative for INRIA

2. eb- nstrumented
an- achine nteractions,
ommunities, and emantics
a joint research team between INRIA Sophia Antipolis – Méditerranée
and I3S (CNRS and University Nice Sophia Antipolis).

3. members
Head (and INRIA contact): Fabien Gandon
Vice Head (and I3S contact): Catherine Faron-Zucker
Researchers:
1.
Michel Buffa, MdC (UNS)
2.
Olivier Corby, CR1 (INRIA)
3.
Alain Giboin, CR1 (INRIA)
4.
Nhan Le Thanh, Pr. (UNS)
5.
Isabelle Mirbel, MdC, HDR (UNS)
6.
Peter Sander, Pr. (UNS)
7.
David Simoncini, ATER (UNS)
8.
Andrea G. B. Tettamanzi, Pr. (UNS)
9.
Serena Villata, RP (INRIA)
Post-doc:
1.
Jodi Schneider(ERCIM)
2.
Elena Cabrio (CORDIS)
Research engineers:
1.
Christophe Desclaux (Boost your code)
2.
Fuqi Song (Inria ADT)
PhD students:
1. Pavel Arapov, 3rd year (EDSTIC-I3S)
2. Amel Benothman, 1st year (I3S)
3. Franck Berthelon, 4th year (UNS-EDSTIC)
4. Ahlem Bouchahda, 3rd year (UNS-SupCom Tunis)
5. Khalil Riad Bouzidi, 3rd year (UNS-CSTB)
6. Papa Fary Diallo, 2nd year (AUF-UGB-INRIA)
7. Amosse Edouard (EDSTIC-I3S)
8. Corentin Follenfant, 3rd year (SAP)
9. Rakebul Hasan, 2nd year (INRIA ANR-Kolflow)
10. Maxime Lefrançois, 3rd year (EDSTIC-INRIA)
11. Nicolas Marie, 3rd year (Bell-ALU, INRIA)
12. Zide Meng, 1st year (INRIA ANR OCTOPUS)
13. Nguyen Thi Hoa Hue, 2nd year (Vietnam-CROUS)
14. Tuan Anh Pham (Vietnam-CampusFrance)
15. Oumy Seye, 2nd year, (INRIA Rose Dieng allocation)
Assistants:
•
Christine Foggia (INRIA)
•
Magali Richir (I3S)

4. research problem
socio-semantic networks: combining formal
semantics and social semantics on the web

5. web landscape and graphs
(meta)data of the relations and the resources of the web
=
web…
+
…sites
=
typed
graphs
+
…social
+
web
(graphs)
…of data
+
networks
(graphs)
+
+
…of services
+
linked data
(graphs)
+…
…semantics
+
workflows
(graphs)
+…
schemas
(graphs)

6. challenge
typed graphs to analyze, model, formalize and implement
social semantic web applications for epistemic communities
 multidisciplinary approach for analyzing and modeling
the many aspects of intertwined information systems
communities of users and their interactions
 formalizing and reasoning on these models using typed graphs
new analysis tools and indicators
new functionalities and better management

7. <background_knowledge>

8. internet
n↔n
classical web
1↔n
wiki, (µ)blog, forum, etc.
n↔n
Ward Cunningham, 94
read-write web

10. social web networks

11. SOCIAL TAGGING
collaboratively create
and manage tags to
annotate and
categorize content

12. a crowd of users creating massive
categorizations
T
U
R

13. semantic web
mentioned by Tim BL
in 1994 at WWW
[Tim Berners-Lee 1994, http://www.w3.org/Talks/WWW94Tim/]

14. SEMANTIC WEB STACK
W3C®

15. A WEB OF
LINKED DATA
W3C®

16. RDF stands for
Resource: pages, images, videos, ...
everything that can have a URI
Description: attributes, features, and
relations of the resources
Framework: model, languages and
syntaxes for these descriptions

17. RDF is a triple model i.e. every
piece of knowledge is broken down into
( subject , predicate , object )

18. doc.html has for author Fabien
and has for theme Music

19. doc.html has for author Fabien
doc.html has for theme Music

20. ( doc.html , author , Fabien )
( doc.html , theme , Music )
( subject , predicate , object )

21. RDFtriples can be seen as arcs
of a graph (vertex,edge,vertex)
(the RDF data model can be seen as a directed labelled multigraph)

22. Fabien
author
doc.html
theme
Music

24. identify what
exists on the
web
http://my-site.fr
identify,
on the web,
what exists
http://animals.org/this-zebra

25. http://ns.inria.fr/fabien.gandon#me
http://inria.fr/schema#author
http://inria.fr/rr/doc.html
http://inria.fr/schema#theme
"Music"

26. principles




use RDF as data format
use URIs as names for things
use HTTP URIs so that people can look up those names
when someone looks up a URI, provide useful information
(RDF, HTML, etc.) using content negotiation
 include links to other URIs so that related things can be
discovered

27. May 2007
April 2008
September 2008
Linking Open Data
March 2009
400
300
200
100
0
10/10/2006 28/04/2007 14/11/2007 01/06/2008 18/12/2008 06/07/2009 22/01/2010 10/08/2010 26/02/2011 14/09/2011 01/04/2012
September 2011
Linking Open Data cloud diagram, by Richard Cyganiak and Anja Jentzsch. http://lod-cloud.net/
September 2010

28. query with SPARQL
SPARQL Protocol and RDF
Query Language

29. e.g. DBpedia

30. opening datathat break the linked nature of the web
silos
created by applications and

31. Ontology
ontologies
->

32. PUBLISHED
SEMANTICS
OF SCHEMAS
W3C®

33. RDFS to declare classes of
resources, properties, and
organize their hierarchy
Document
creator
author
Report
Document
Person

34. OWL in one…

algebraic properties
!
union
disjunction
intersection
complement
restriction
1..1
disjoint properties
1..1
! qualified cardinality
individual prop. neg


chained prop.
cardinality
 equivalence
enumeration
[>18] value restrict.
disjoint union

keys
…

35. SKOS
knowledge
thesauri,
classifications,
subjects,
taxonomies,
folksonomies,
... controlled
vocabulary
35

36. labels
natural language expressions to refer to concepts
inria:CorporateSemanticWeb
skos:prefLabel "corporate semantic web"@en;
skos:prefLabel "web sémantique d'entreprise"@fr;
skos:altLabel "corporate SW"@en;
skos:altLabel "CSW"@en;
skos:hiddenLabel "web semantique d'entreprise"@fr.
36

37. relations
between concepts
inria:CorporateSemanticWeb
skos:broader w3c:SemanticWeb;
skos:narrower inria:CorporateSemanticWiki;
skos:related inria:KnowledgeManagement.

38. </background_knowledge>

39. http://wimmics.inria.fr

41. [Limpens, et al.]
e.g. structuring folksonomy
web 2.0
flat folksonomies
thesaurus
pollutant
energy
related
related
?
pollution
has narrower
soil pollution
SKOS

42. [Limpens, et al.]
e.g. combining metric spaces
edition distances
Monge-Elkan Soundex, JaroWinkler,
asymmetry Monge-Elkan Qgram
contextual metric
cosinus vector of co-occurring tags
tag1  tag 2
cos tag1 , tag 2 
tag1  tag 2


social metrics
inclusion of communities of interest
pollution
environment

43. e.g. ademe TheseNet
83 027 relations / 9 037 tags
 68 633 related
 11 254 hyponyms
 3 193 spelling variants

44. evolution of the place of humans in
complex web applications
=
user
=
data
=
processor

45. [Limpens, et al.]
e.g. search & feedback

47. [Erétéo, et al.]
e.g. typing sociograms
Fabien
Michel
Guillaume
Marco
Rémi

din ( p )  x ; rel( x , p )

din (Guillaume)  4
Nicolas
social network analysis
Man
creator
type
author
Fabien
type
Person
creator
Person
sub property
sub class
doc.html
title
Semantic web is not antisocial
author
Man
semantic web

48. Fabien
Mylène
Gérard
knows
(guillaume)=3
guillaume
d(guillaume)=5
<family>
colleague
Michel
sibling
sister
brother
parent
father mother
Yvonne

49. [Erétéo, et al.]
e.g. parameterized analysis
SNA indices
(G)
actor
type
nb
actor
nbrel (G)
subject
nbrel (G)
SPARQL formal definition
g
nbrel (b, from, to)
select ?from ?to ?b count($path) as ?count
where{
?from sa (param[rel])*::$path ?to
graph $path{?b param[rel] ?j}
filter(?from != ?b)
optional { ?from param[rel]::$p ?to }
filter(!bound($p))
}group by ?from ?to ?b
c
Crel ( y)
select distinct ?y ?to pathLength($path) as
?length (1/sum(?length)) as ?centrality
where{
?y s (param[rel])*::$path ?to
}group by ?y
select ?from ?to ?b
(count($path)/count($path2)) as ?betweenness
where{
?from sa (param[rel])*::$path ?to
graph $path{?b param[rel] ?j}
filter(?from != ?b)
optional { ?from param[rel]::$p ?to }
filter(!bound($p))
?from sa (param[rel])*::$path2 ?to
}group by ?from ?to ?b
select merge count(?x) as ?nbactor from <G>
where{
?x rdf:type param[type]
}
select merge count(?x) as ?nbactors from <G>
where{
{?x param[rel] ?y}
UNION{?y param[rel] ?x}
}
select merge count(?x) as ?nbsubj from <G>
where{
?x param[rel] ?y
}
object
nbrel (G)
select merge count(?y) as ?nbobj from <G>
where{
?x param[rel] ?y
}
relation
nbrel (G)
select cardinality(?p) as ?card from <G>
where {
{ ?p rdf:type rdf:Property
filter(?p ^ param[rel]) }
UNION
{ ?p rdfs:subPropertyOf ?parent
filter(?parent ^ param[rel]) }
}
select ?x ?y from <G> where {
?x param[rel] ?y
}group by any
select ?y count(?x) as ?degree where {
{?x (param[rel])*::$path ?y
filter(pathLength($path) <= param[dist])}
UNION
{?y param[rel]::$path ?x
filter(pathLength($path) <= param[dist])}
}group by ?y
select ?y count(?x) as ?indegree where{
?x (param[rel])*::$path ?y
filter(pathLength($path) <= param[dist])
}group by ?y
select ?x count(?y) as ?outdegree where {
?x (param[rel])*::$path ?y
filter(pathLength($path) <= param[dist])
}group by ?x
Brel (b, from, to)
Comprel (G)
Drel ,dist ( y)
in
Drel ,dist ( y)
out
Drel ,dist ( y)
g rel ( from, to)
select ?from ?to $path pathLength($path) as
?length where{
?from sa (param[rel])*::$path ?to
}group by ?from ?to
Diamrel (G)
select pathLength($path) as ?length from <G>
where {
?y s (param[rel])*::$path ?to
}order by desc(?length)
limit 1
select ?from ?to count($path) as ?count
where{
?from sa (param[rel])*::$path ?to
}group by ?from ?to
g
nbrel ( from, to)
nbgrel  (b, x, y)
B rel  b, x, y  
nbgrel  ( x, y )
:=
select ?from ?to ?b
(count($path)/count($path2)) as
?betweenness where{
?from sa (param[rel])*::$path ?to
graph $path{?b param[rel] ?j}
filter(?from != ?b)
optional { ?from param[rel]::$p ?to}
filter(!bound($p))
?from sa (param[rel])*::$path2 ?to
}group by ?from ?to ?b

50. [Erétéo, et al.]
ipernity.com dataset in RDF
61 937 actors & 494 510 relationships
–18 771 family links between 8 047 actors
–136 311 friend links implicating 17 441 actors
–339 428 favorite links for 61 425 actors, etc.
existence of a largest component in all sub networks
"the effectiveness of the social network at doing its job"
[Newman 2003]
70000
60000
50000
40000
30000
20000
10000
0
know s
favorite
friend
fam ily
m essage
num ber actors
size largest com ponent
com m ent

51. typed centrality:
different key actors for
different kinds of links

53. detecting AND labeling communities
?
?

54. [Eretéo, et al.]
e.g. semantic propagation
from RAK/LP to SemTagP
rugby, foot
hockey sel, eau
poivre, vin
foot, ciné
moutarde
sport
sport
condiment
condiment
sport condiment

55. applied to Ademe Ph.D. network
 1 853 agents
1 597 academic supervisors
256 ADEME engineers.
 13 982 relationships
10 246 rel:worksWith
3 736 rel:colleagueOf
 6 583 tags
 3 570 skos:narrower
relations between 2 785 tags

56. e.g. Ademe
1 pollution ; 2 développent durable ;
3 énergie ; 4 chimie ; 5 pollution de l’air ;
6 métaux ; 7 biomasse ; 8 déchets.
[Erétéo, et al.]

59. towards rich webmarks

60. [Delaforge, et al.]
Fresnel lenses to adapt results

61. toward social webmarks

62. socially defined online identities

63. [Delaforge, et al.]

64. [Delaforge, et al.]

65. [Delaforge, et al.]
search & navigation in info. networks

66. Semantic Web plugin for Gephi

67. activity flow and notification

68. web-scraping: archiving and integrating

69. [Buffa, et al.]
create dynamic reports in the wiki

71. gave birth to …
« Unveil the semantic imprint from within your Community »
cooperative society of social semantic web specialists
industrialization and maintenance of research results
developing webmarks as linked semantic traces
exhange
contributes
interested
in
Xxxxxx
xxxxxx
xxxxxx
Xxxxxx
xxxxxx
xxxxxx
link & enrich
contributes
exchange
interested in
Xxxxx
xxx
xxxxx
xxxx
Xxxxx
xxxx
xxxx
contribute
Xxxxxx
xxxxxx
xxxxxxXxxxxx
xxxxxx
xxxxxx
interested in
downloaded
Xxxxxx
xxxxxx
xxxxxxXxxxxx
xxxxxx
xxxxxx
analyse & assist
Xxxxx
x
xxxxx

73. going
mobile

74. mobile access to web of data
&
Mobile
&
Web of
Data
&
Context
Interaction
prissma
[Costabello et al.]

75. Context-aware Adaptation for Linked Data?
?
wimmics.inria.fr/projects/prissma
[Costabello et al.]

76. fault-tolerant matching with graph-edit distances
prissma:environment
:ActualCtx
prissma:poi
:actualPOI
:actualEnv
C=0
geo:lat
prissma:radius
geo:long
10
48.843453
2.32434
✓
C=0.34
C=0 ✓
✓
:museumGeo
prissma:Context
0
prissma:radius
geo:lat
geo:lon
200
48.86034
-2.337599
1
prissma:Environment
2
C=0.17
✓
{ 3, 1, 2, { pr i ssm poi } }
a:
C=0.09
✓
:atTheMuseum
{ 4, 0, 3, { pr i ssm envi r onm
a:
ent } }
[Costabello et al.]

77. examples
PRISSMA Browser for Android
Smartphone, user walking
in museum town.
Tablet, user at home.
[Costabello et al.]

78. when the linklinking data to create knwoledge
makes sense

79. contextual notification
propagation of interests for suggestion
[Marie, et al.]
0,4
0,9
0,7
𝒂 𝒊, 𝒏 + 𝟏, 𝒕 =
𝒘 𝒔 ∗ 𝐬(𝐢, 𝐧 + 𝟏, 𝐭) +
𝒘𝒔 +
𝐣,𝒆 𝐢,𝒋
𝐣,𝒆 𝐢,𝒋
𝒘 𝒑 𝐥 𝒆 𝐢,𝒋 , 𝒕 ∗ 𝐚(𝐣, 𝐧, 𝐭)
𝒘 𝒑 𝒆 𝐢,𝒋

80. Discovery Hub
1
2
3
DBpedia:
Claude Monet
4
Results identification
Ranking
Sorting/categorization
Explanation

81. Discovery Hub (Inria, Alcatel Bell Lucent)

82. ELIZA… talking to machines

83. Who is starring in
Batman Begins?
EAT and NE recognition:
Stanford NER+ DBpedia
Relational Patterns extraction
Query pattern
Pattern repository
owner(Thing, Thing)
[Person] is starring
in [Movie]?
[R:Person] purchased the [D:Thing]
[D:Thing] owner [R:Thing]
[D:Thing] was bought by [R:Thing]
question
answering
over
linked
data
[Cabrio, et al.]
select * where {
dbpr:Batman_Begins dbp:starring ?v
OPTIONAL {?v rdfs:label ?l
filter(lang(?l)="en")} }
starring(Work, Person)
.
[D:Work], played by [R:Person]
[D:Work] stars [R:Person]
[D:Work] film stars [R:Person]
ENTAILMENT ENGINE/
SIMILARITY ALGORITHM
QALD-2 Open Challenge:
Christian Bale, Michael Caine, ...

84. QAKIS demo

85. EMOTION
detection
transfer
adaptation
www

86. ns.inria.fr/emoca
[Berthelon, et al.]

88. e.g. DBpedia
[Cojan, et al.]

89. publication
Datalift process
demo
• on click setup
• raw data import
• RDF transformation
• Web publication
• online querying

91. [Corby, et al.]
corese/kgram
• Semantic Web Factory: RDF/S, SPARQL 1.1
Query & Updade, Inference Rules
• Open Source CeCILL-C
• Knowledge Graph Abstract Machine
with 3 Proxies (Producer, Matcher, Evaluator)
3 ANR, 2 RNRT, 1 region project, 4 european project,
4 industry grants, 10 academic grants,
>30 applications, 23 PhD, 9 edu. Institutions, etc.

92. left(x,y)
left(y,z)
right(z,v)
right(z,u)
right(u,v)
left(x,?p) left(?p,z)

left
x
*
left
z
y
left
left
x
z
right
right
u
v

93. vehicle
O
vehicle
GR
car(x)vehicle(x)
car
car
GF
FOmodulo an ontologyGF  GR
R
mapping

94. CORESE/ KGRAM
[Corby, et al.]

95. my watch has only one hand,
it is not broken, it is a feature.
why approximation is interesting

96. e.g. controlled approximation
vehicle
car
O
car(x) … truck(x)
car
truck
truck
t1(x)t2(x)  d(t1,t2)< threshold

(t1 , t 2 )  H c let dist (t1 , t 2 )  min t t1 ,t t2  l H c (t1 , t )  l H c (t 2 , t )
2
(t1 , t 2 )  H c ; t1  t 2 let l H c (t1 , t 2 )  t t ,t
2
1 2
 1 
,t  t1   depth ( t ) 
2



97. e.g. approximated search

98. Plugin Gephi
[Demairy, et al.]

99. RDF
RDF
peer
servers
web service
web service
web service
web application
RDF
RDF
SPARQL

100. inductive index creation for a triple store
[Basse, et al.]
• characterize distributed RDF sources
• incremental index generation and maintenance

101. [Gaignard, et al.]
federating KGram

103. control

104. [Villata, et al.]
socio-semantic access control
User
e.g. only my colleagues
working on the same subject
ASK{ ?res dcterms:creator ?prov .
?prov rel:hasColleague ?user .
?prov foaf:interestedBy ?topic .
?user foaf:interestedBy ?topic }

105. Context-Aware Access Control
Model [Villata, Costabello, et al.]
s4ac:
DisjunctiveACS
subClassOf
hasAccessPrivilege
hasAccessConditionSet
subClassOf
ConjunctiveACS
appliesTo
AccessPolicy
AccessPrivilege
AccessConditionSet
hasAccessCondition
AccessCondition
hasQueryAsk
Device
device
hasContext
Context
User
user
environment
Environment
105

106. Context-aware Access Control for Linked Data
Shi3ld Access Control Manager
SELECT …
WHERE {…}
GET /data/resource HTTP/1.1
Host: example.org
Authorization: ...
wimmics.inria.fr/projects/shi3ld

107. toward the « oh yeah? » button

108. representing query and reasoning
workflows [Hasan et al.]
• Ratio4TA*, a lightweight
vocabulary for encoding
justifications.
• A specialization of the W3C
PROV ontology
*
http://ns.inria.fr/ratio4ta/

109. [Hasan et al.]
overwhelming…
evaluating quality
of summaries

110. doggy-bag
of the talk

111. web 1, 2

112. web 1, 2, 3
person
price
homepage?
convert?
more info?

113. wikipedia editions by users…
3500000
3000000
2500000
2000000
1500000
1000000
500000
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
3500000
without bots…
3000000
2500000
2000000
1500000
1000000
500000
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

114. one web…
person
document
informal
formal
program
data
metadata
usage
representation

115. he who controls metadata, controls the web
and through the world-wide web many things in our world.
fabien, gandon, @fabien_gandon, http://fabien.info