wimmics research team 2015 activity report

WIMMICSWeb-instrumented man-machine interactions, communities and semantics

Fabien GANDONhttp://fabien.info

SOPHIA TEAM

Inria

CNRS

University of Nice

Inria Lille - Nord Europe (2008)

Inria Saclay – Ile-de-France

(2008)

Inria Nancy – Grand Est

(1986)

Inria Grenoble – Rhône-

Alpes (1992)

Inria Sophia Antipolis Méditerranée (1983)

Inria Bordeaux

Sud-Ouest (2008)

Inria Rennes

Bretagne

Atlantique

(1980)

Inria Paris-Rocquencourt

(1967)

Montpellier

Lyon

Nantes

Strasbourg

Center

Branch

Pau

I3S

CHALLENGEto bridge social semantics andformal semantics on the Web

MULTI-DISCIPLINARY TEAM

50 members (2015)

14 nationalities

1 DR, 3 Professors

3CR, 4 Assistant professors

1 SRP

DR/Professors: Fabien GANDON, Inria, AI, KR, Semantic Web, Social Web Nhan LE THANH, UNS, Logics, KR, Emotions Peter SANDER, UNS, Web, Emotions Andrea TETTAMANZI, UNS, AI, Logics, Agents,

CR/Assistant Professors: Michel BUFFA, UNS, Web, Social Media Elena CABRIO, UNS, NLP, KR, Linguistics Olivier CORBY, Inria, KR, AI, Sem. Web, Programming, Graphs Catherine FARON-ZUCKER, UNS, KR, AI, Semantic Web, Graphs Alain GIBOIN, Inria, Interaction Design, KE, User & Task models Isabelle MIRBEL, UNS, Requirements, Communities Serena VILLATA, CNRS, AI, Argumentation, Licenses, Rights

Inria Starting Position: Alexandre MONNIN, Philosophy, Web

RESEARCH CHALLENGES

1. user & interaction design

2. communities & social networks

3. linked data & semantic Web

4. reasoning & analyzing

RESEARCH CHALLENGES





How do we improve our interactions with asemantic and social Web ?

• capture and model the users' characteristics?

• represent and reason with the users’ profiles?

• adapt the system behaviors as a result?

• design new interaction means?

• evaluate the quality of the interaction designed?

RESEARCH CHALLENGES





How can we manage the collective activity on socialmedia?

• analyze the social interaction practices and the

structures in which these practices take place?

• capture the social interactions and structures?

• formalize the models of these social constructs?

• analyze & reason on these models of social activity?

RESEARCH CHALLENGES





What are the needed schemas and extensions ofthe semantic Web formalisms for our models?

• formalisms best suited for the models of the

challenges 1 & 2 ?

• limitations and extensions of existing formalisms?

• missing schemas, ontologies, vocabularies?

• links and combinations of existing formalisms?

RESEARCH CHALLENGES





What are the algorithms required to analyze andreason on the heterogeneous graphs we obtained?

• analyze graphs of different types and their

interactions?

• support different graph life-cycles, calculations and

characteristics?

• assist different tasks of our users?

• design the Web architecture to deploy this?

METHODS AND TOOLS





G2 H2

G1 H1

<Gn Hn

METHODS AND TOOLS





• user models, personas, emotion capture

• mockups, evaluation campaigns

• KB interaction (context, Q&A, exploration, …)

G2 H2

G1 H1

<Gn Hn

METHODS AND TOOLS








• collective personas, coordinative artifacts

• community detection, labelling

• argumentation theory, sentiment analysis

G2 H2

G1 H1

<Gn Hn

METHODS AND TOOLS











• ontology-based knowledge representation

• formalisms: typed graphs, uncertainty

• knowledge extraction, data translation

G2 H2

G1 H1

<Gn Hn

METHODS AND TOOLS











• ontology-based knowledge representation

• formalisms: typed graphs, uncertainty

• knowledge extraction, data translation

• graph querying, reasoning, transforming

• induction, propagation, approximation

• explanation, tracing, control, licensing, trust

RESULTS 2013-2015





• e.g. DiscoveryHub, PRISSMA, QAKiS

[IJCAI, WI, Semantics, ESWC, Hypertext, IJSWIS, …]

• e.g. OCKTOPUS, WikiNEXT, SEEMPAD

[AI, CHI, ECAI, WI, COMMA, ASONAM, IAT, SAC, …]

• e.g. L4LOD, S4AC, Ratio4TA, Dbpedia.fr

[ECAI, IJCAI, LREC, AAMAS, RuleML, WebIST, …]

• e.g. CORESE, Licentia, DiscoveryHub, PRISSMA

[ISWC, EKAW, KCAP, WI, ESWC, Hypertext, ICAIL, …]

SEARCHING exploratory search

question-answering


question-answering

DBPEDIA.FR (extraction, end-point)180 000 000 triples

modelsWeb architecture


question-answering


DISCOVERYHUB.CO

semantic spreadingactivation

new evaluation protocol


question-answering


DISCOVERYHUB.CO

QAKiS.ORG

semantic spreadingactivation

new evaluation protocol

[D:Work], played by [R:Person]

[D:Work] stars [R:Person]

[D:Work] film stars [R:Person]

starring(Work, Person)

linguistic relationalpattern extraction

named entity recognitionsimilarity based SPARQLgeneration

select * where {

dbpr:Batman_Begins dbp:starring ?v .

OPTIONAL {?v rdfs:label ?l

filter(lang(?l)="en")} }

MODELING USERS individual context

social structures


social structures

PRISSMAprissma:Context

0 48.86034

-2.337599

200

geo:latgeo:lon

prissma:radius

1

:museumGeo

prissma:Environment

2

{ 3, 1, 2, { pr i ssma: poi } }

{ 4, 0, 3, { pr i ssma: envi r onment } }

:atTheMuseum

error tolerant graphedit distance

contextontology


social structures


0 48.86034

-2.337599

200

geo:latgeo:lon

prissma:radius

1

:museumGeo

prissma:Environment

2

{ 3, 1, 2, { pr i ssma: poi } }


:atTheMuseum


contextontology

OCKTOPUS

tag, topic, userdistribution

tag and folksonomyrestructuring with

prefix trees


social structures


0 48.86034

-2.337599

200

geo:latgeo:lon

prissma:radius

1

:museumGeo

prissma:Environment

2

{ 3, 1, 2, { pr i ssma: poi } }


:atTheMuseum


contextontology

OCKTOPUS

tag, topic, userdistribution

tag and folksonomyrestructuring with

prefix trees

EMOCA&SEEMPAD

emotion detection & annotation

QUERY & INFER graph rules and queries

deontic reasoning

induction


deontic reasoning

induction

CORESE

& G2 H2

&G1 H1

<Gn Hn

abstract graph machineSTTL


deontic reasoning

induction

CORESE

& G2 H2

&G1 H1

<Gn Hn

RATIO4TA

predict &explain



deontic reasoning

induction

CORESE

INDUCTION

& G2 H2

&G1 H1

<Gn Hn

RATIO4TA

predict &explain

find missingknowledge



deontic reasoning

induction

CORESE

LICENTIA INDUCTION

& G2 H2

&G1 H1

<Gn Hn

RATIO4TA

predict &explain

find missingknowledge

license compatibilityand composition


ACHIEVEMENTS

• publications

• awards

• visibility

• 9 Ph.D. theses defended

• multidisciplinary publication

• best paper awards: ESWC 2013 ; EEE/WIC/ACM IAT 2014 ; IEEE Cognitive Infocommunications ; IEEE BDAS 2014 ; IC2015

• PC of major journals and conferences

• General Chair ESWC 2015, PC chair ESWC 2014, ...

COLLABORATIONS

• collaborative projects (7)

• industrial contracts (6)

• scholarships (8)

+ MIREL

DIFFUSION / TRANSFER

• education, training

• industrial transfer

• two MOOCS (HTML5 EN 100K, Semantic Web FR 4K)

• important teaching activity

• W3C participation (AC Rep, several WG)

• 2 Carnot projects with SMEs

• first ANR LabCom on digital sciencesjoint Lab Wimmics-Viseo

• spin-off to industrialize research resultsone of the first software cooperatives

FUTURE

• Linked Data & Web diversity

• Artificial Web intelligence

• Human-Data Web Interactions

• Web-augmented interactions

• coupling AI and distributed AI with the Web

• knowledge representation & extraction on the Web

• incremental & linked formalizations

• argumentation mining & theory

• textual entailment & LOD-based NLP

• participatory design and evaluation

• natural language question-answering and dialogue

• visual analytics for linked data

• architecture & philosophy of the Web

WIMMICSbridging social semantics and formal semantics on the Web.

epistemic communitieslinked data

usages and introspection

contributions and traces

wimmics research team 2015 activity report

Science