bridging formal semantics and social semantics on the web

  • 1,595 views
Uploaded on

keynote for SMAP 2013

keynote for SMAP 2013

More in: Technology , Education
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
  • Nope, many PhD results are missing and the idea was not to talk about the future work of the team but about the results corresponding to the topics of SMAP2013
    Are you sure you want to
    Your message goes here
  • So you bring all the PhD work into one big slideset? We can see the overall story but we are missing where this is going ...
    Are you sure you want to
    Your message goes here
No Downloads

Views

Total Views
1,595
On Slideshare
0
From Embeds
0
Number of Embeds
4

Actions

Shares
Downloads
69
Comments
2
Likes
4

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 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
  • 9. social web networks
  • 10. SOCIAL TAGGING collaboratively create and manage tags to annotate and categorize content
  • 11. a crowd of users creating massive categorizations T U R
  • 12. semantic web mentioned by Tim BL in 1994 at WWW [Tim Berners-Lee 1994, http://www.w3.org/Talks/WWW94Tim/]
  • 13. SEMANTIC WEB STACK W3C®
  • 14. A WEB OF LINKED DATA W3C®
  • 15. 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
  • 16. RDF is a triple model i.e. every piece of knowledge is broken down into ( subject , predicate , object )
  • 17. doc.html has for author Fabien and has for theme Music
  • 18. doc.html has for author Fabien doc.html has for theme Music
  • 19. ( doc.html , author , Fabien ) ( doc.html , theme , Music ) ( subject , predicate , object )
  • 20. RDFtriples can be seen as arcs of a graph (vertex,edge,vertex) (the RDF data model can be seen as a directed labelled multigraph)
  • 21. Fabien author doc.html theme Music
  • 22. identify what exists on the web http://my-site.fr identify, on the web, what exists http://animals.org/this-zebra
  • 23. http://ns.inria.fr/fabien.gandon#me http://inria.fr/schema#author http://inria.fr/rr/doc.html http://inria.fr/schema#theme "Music"
  • 24. 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
  • 25. 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
  • 26. query with SPARQL SPARQL Protocol and RDF Query Language
  • 27. e.g. DBpedia
  • 28. opening datathat break the linked nature of the web silos created by applications and
  • 29. Ontology ontologies ->
  • 30. PUBLISHED SEMANTICS OF SCHEMAS W3C®
  • 31. RDFS to declare classes of resources, properties, and organize their hierarchy Document creator author Report Document Person
  • 32. 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 …
  • 33. SKOS knowledge thesauri, classifications, subjects, taxonomies, folksonomies, ... controlled vocabulary 35
  • 34. 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
  • 35. relations between concepts inria:CorporateSemanticWeb skos:broader w3c:SemanticWeb; skos:narrower inria:CorporateSemanticWiki; skos:related inria:KnowledgeManagement.
  • 36. </background_knowledge>
  • 37. http://wimmics.inria.fr
  • 38. [Limpens, et al.] e.g. structuring folksonomy web 2.0 flat folksonomies thesaurus pollutant energy related related ? pollution has narrower soil pollution SKOS
  • 39. [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
  • 40. e.g. ademe TheseNet 83 027 relations / 9 037 tags  68 633 related  11 254 hyponyms  3 193 spelling variants
  • 41. evolution of the place of humans in complex web applications = user = data = processor
  • 42. [Limpens, et al.] e.g. search & feedback
  • 43. [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
  • 44. Fabien Mylène Gérard knows (guillaume)=3 guillaume d(guillaume)=5 <family> colleague Michel sibling sister brother parent father mother Yvonne
  • 45. [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
  • 46. [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
  • 47. typed centrality: different key actors for different kinds of links
  • 48. detecting AND labeling communities ? ?
  • 49. [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
  • 50. 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
  • 51. 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.]
  • 52. towards rich webmarks
  • 53. [Delaforge, et al.] Fresnel lenses to adapt results
  • 54. toward social webmarks
  • 55. socially defined online identities
  • 56. [Delaforge, et al.]
  • 57. [Delaforge, et al.]
  • 58. [Delaforge, et al.] search & navigation in info. networks
  • 59. Semantic Web plugin for Gephi
  • 60. activity flow and notification
  • 61. web-scraping: archiving and integrating
  • 62. [Buffa, et al.] create dynamic reports in the wiki
  • 63. 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
  • 64. going mobile
  • 65. mobile access to web of data & Mobile & Web of Data & Context Interaction prissma [Costabello et al.]
  • 66. Context-aware Adaptation for Linked Data? ? wimmics.inria.fr/projects/prissma [Costabello et al.]
  • 67. 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.]
  • 68. examples PRISSMA Browser for Android Smartphone, user walking in museum town. Tablet, user at home. [Costabello et al.]
  • 69. when the linklinking data to create knwoledge makes sense
  • 70. 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
  • 71. Discovery Hub 1 2 3 DBpedia: Claude Monet 4 Results identification Ranking Sorting/categorization Explanation
  • 72. Discovery Hub (Inria, Alcatel Bell Lucent)
  • 73. ELIZA… talking to machines
  • 74. 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, ...
  • 75. QAKIS demo
  • 76. EMOTION detection transfer adaptation www
  • 77. ns.inria.fr/emoca [Berthelon, et al.]
  • 78. e.g. DBpedia [Cojan, et al.]
  • 79. publication Datalift process demo • on click setup • raw data import • RDF transformation • Web publication • online querying
  • 80. [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.
  • 81. 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
  • 82. vehicle O vehicle GR car(x)vehicle(x) car car GF FOmodulo an ontologyGF  GR R mapping
  • 83. CORESE/ KGRAM [Corby, et al.]
  • 84. my watch has only one hand, it is not broken, it is a feature. why approximation is interesting
  • 85. 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  
  • 86. e.g. approximated search
  • 87. Plugin Gephi [Demairy, et al.]
  • 88. RDF RDF peer servers web service web service web service web application RDF RDF SPARQL
  • 89. inductive index creation for a triple store [Basse, et al.] • characterize distributed RDF sources • incremental index generation and maintenance
  • 90. [Gaignard, et al.] federating KGram
  • 91. control
  • 92. [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 }
  • 93. 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
  • 94. 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
  • 95. toward the « oh yeah? » button
  • 96. 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/
  • 97. [Hasan et al.] overwhelming… evaluating quality of summaries
  • 98. doggy-bag of the talk
  • 99. web 1, 2
  • 100. web 1, 2, 3 person price homepage? convert? more info?
  • 101. 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
  • 102. one web… person document informal formal program data metadata usage representation
  • 103. 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