EKAW 2022 keynote by Fabien GANDON: "a shift in our research focus: from knowledge acquisition to knowledge augmentation"
While EKAW started in 1987 as the European Knowledge Acquisition Workshop, in 2000 it transformed into a conference where we advance knowledge engineering and modelling in general. At the time, this transition also echoed shifts of focus such as moving from the paradigm of expert systems to the more encompassing one of knowledge-based systems. Nowadays, with the current strong interest for knowledge graphs, it is important again to reaffirm that our ultimate goal is not the acquisition of bigger siloed knowledge bases but to support knowledge requisition by and for all kinds of intelligence. Knowledge without intelligence is a highly perishable resource. Intelligence without knowledge is doomed to stagnation. We will defend that intelligence and knowledge, and their evolutions, have to be considered jointly and that the Web is providing a social hypermedia to link them in all their forms. Using examples from several projects, we will suggest that, just like intelligence augmentation and amplification insist on putting humans at the center of the design of artificial intelligence methods, we should think in terms of knowledge augmentation and amplification and we should design a knowledge web to be an enabler of the futures we want.
a shift in our research focus: from knowledge acquisition to knowledge augmentation
1. a shift in our research focus:
from knowledge acquisition
to knowledge augmentation
Fabien Gandon – Keynote EKAW 2022
@fabien_gandon
http://fabien.info
2. WIMMICS TEAM
▪ Inria
▪ CNRS
▪ University Côte D’Azur (UCA)
I3S
Web-Instrumented Man-Machine Interactions,
Communities and Semantics
3. MULTI-DISCIPLINARY TEAM
▪ 35~55 members
▪ ~15 nationalities
▪ 2 DR, 5 Professors
▪ 1CR, 1 Assistant professors
DR / Tenured Senior Researchers / Professors:
▪ Michel BUFFA, UCA, Web, Social Media, Web Audio/Music, K. Graphs
▪ Elena CABRIO, UCA, NLP, KR, Linguistics, Q&A, Text Mining, K. Graphs
▪ Catherine FARON, UCA, KR, AI, Semantic Web, K. Graphs
▪ Fabien GANDON, Inria, AI, KRR, Semantic Web, Social Web, K. Graphs
▪ Andrea TETTAMANZI, UCA, AI, Logics, Evo, Learning, Agents, K. Graphs
▪ Serena VILLATA, CNRS, AI, Argumentation, Licenses, Rights, K. Graphs
▪ Marco WINCKLER, UCA, Human-Computer Interaction, Web, K. Graphs
CR / Tenured Junior Researchers / Assistant Professors:
▪ Damien GRAUX, Inria, Linked Data, Sem. Web, Querying, K. Graphs
▪ Aline MENIN, UCA, Human-Computer Interaction, Web, K. Graphs
Research engineer: Franck MICHEL, CNRS, Linked Data, Integration, DB, K. Graphs
3IA Associate Professor: Amaya Nogales Gómez, UCA, Fairness, Bias, Opti. , ML
Retired / About to retire:
▪ Olivier CORBY, Inria, CR, KR, AI, Sem. Web, Programming, K. Graphs
▪ Alain GIBOIN, Inria, CR, Interaction Design, KE, User & Task, K. Graphs
▪ Nhan LE THANH, UCA, PR, Logics, KR, Emotions, Workflows, K. Graphs
▪ Peter SANDER, UCA, PR, Web, Emotions
External:
▪ Andrei Ciortea (University of St. Gallen) Agents, WoT, Sem. Web, K. Graphs
▪ Nicolas DELAFORGE (Mnemotix) Sem. Web, KM, Integration, K. Graphs
▪ Freddy LECUE (JP Morgan, NY) AI, Logics, Mining, Big Data, S. Web , K. Graphs
▪ Ștefan SARKADI (King's College, London) AI, Agent Simulation, Deceptive AI
4. EKAW (European Knowledge Acquisition Workshop) since 1987
Analogue to the KAW (Knowledge Acquisition Workshop) in North America
and PKAW in the Asian-Pacific area
From the (European) Knowledge
Acquisition Workshop to the
International Conference on
Knowledge Engineering and
Knowledge Management
6. position/history paper in 2012
“Clearly, the shift from early works comes from the amount of data available (…). The
initial problem of KA of having too few sources has shifted towards the problem of too
many sources, too big, too diverse, too dynamic, too distributed, etc. New challenges
arise in addressing data and schema heterogeneity and provenance, incoherence, and
social life cycles. Old challenges are increased tenfold by the scale and complexity of
the web (acquisition, formalization, evaluation, evolution, etc.).”[Aussenac-Gilles & Gandon, 2012]
10. to google: transitive verb that means using the Google search engine to
obtain information on something or somebody on the World Wide Web.
Nominal Forms
Infinitive: to google
Participle: googled
Gerund: googling
Indicative
Present
I google
you google
he googles
we google
you google
they google
Perfect
I have googled
you have googled
he has googled
we have googled
you have googled
they have googled
Past
I googled
you googled
he googled
we googled
you googled
they googled
Pluperfect
I had googled
you had googled
he had googled
we had googled
you had googled
they had googled
Future
I will google
you will google
he will google
we will google
you will google
they will google
Future perfect
I will have googled
you will have googled
he will have googled
we will have googled
you will have googled
they will have googled
Subjunctive
Present
I google
you google
he google
we google
you google
they google
Perfect
I have googled
you have googled
he have googled
we have googled
you have googled
they have googled
Imperfect
I googled
you googled
he googled
we googled
you googled
they googled
Pluperfect
I had googled
you had googled
he had googled
we had googled
you had googled
they had googled
Conditional
Present
I would google
you would google
he would google
we would google
you would google
they would google
Perfect
I would have googled
you would have googled
he would have googled
we would have googled
you would have googled
they would have googled
Imperative
you google
we Let´s google
you google
Progressive (Continuous)
Forms
Indicative
Present
I am googling
you are googling
he is googling
we are googling
you are googling
they are googling
Perfect
I have been googling
you have been googling
he has been googling
we have been googling
you have been googling
they have been googling
Past
I was googling
you were googling
he was googling
we were googling
you were googling
they were googling
Pluperfect
I had been googling
you had been googling
he had been googling
we had been googling
you had been googling
they had been googling
Future
I will be googling
you will be googling
he will be googling
we will be googling
you will be googling
they will be googling
Future perfect
I will have been googling
you will have been googling
he will have been googling
we will have been googling
you will have been googling
they will have been googling
Conditional
Present
I would be googling
you would be googling
he would be googling
we would be googling
you would be googling
they would be googling
Perfect
I would have been googling
you would have been googling
he would have been googling
we would have been googling
you would have been googling
they would have been googling
12. URI, IRI, URL, HTTP URI
DATA AND SCHEMATA STANDARDS ON THE WEB
JSON
RDF
JSON LD
N-Triple
N-Quad
Turtle/N3
TriG
RDFS
OWL
SPARQL
XML
HTML
RDF XML
HTTP
Linked Data
CSV-LD R2RML
GRDDL
RDFa
SHACL
LDP
13. document and data growth on the Web
# web servers
0
500
1000
1500
# linked open datasets on the Web
14. “What we learned from the WWW (…) is that typically people don’t want
machines to behave like experts; they want to have access to
information so they themselves can exhibit expert performance at just
the right time.”[Allemang, Hendler, Gandon, 2020]
15. initial p sition
Knowledge without intelligence is a highly perishable resource.
Intelligence without knowledge is doomed to stagnation.
Intelligence, knowledge and their evolutions, have to be considered jointly
16. AI for Artificial Intelligence (McCarthy et al., 1955)
IA for Intelligence Amplification (Ashby, 1956) and
Intelligence Augmentation (Engelbart, 1962)
should our focus be shifting to knowledge
augmentation & amplification ?
AI & IA
17. our ultimate goal should be our knowledge
augmentation
knowledge knowledge
21. PUBLISHING
▪ Dbpedia.fr since 2012
▪ extract data (content, activity…)
▪ provide them as linked data
DBPEDIA.FR (extraction, end-point)
650 000 000 triples
[Ringwald, Cojan, Boyer et al.]
22. EXTRACTED
entire edition history as
linked open data
1.9 billion triples describing the 107 million revisions since the first page was created
<http://fr.wikipedia.org/wiki/Victor_Hugo> a prov:Revision ;
dc:subject <http://fr.dbpedia.org/resource/Victor_Hugo> ;
swp:isVersion "3496"^^xsd:integer ;
dc:created "2002-06-06T08:48:32"^^xsd:dateTime ;
dc:modified "2015-10-15T14:17:02"^^xsd:dateTime ;
dbfr:uniqueContributorNb 1295 ;
(...)
dbfr:revPerYear [ dc:date "2015"^^xsd:gYear ; rdf:value
"79"^^xsd:integer ] ;
dbfr:revPerMonth [ dc:date "06/2002"^^xsd:gYearMonth ;
rdf:value "3"^^xsd:integer ] ;
(...)
dbfr:averageSizePerYear [ dc:date "2015"^^xsd:gYear ;
rdf:value "154110.18"^^xsd:float
] ;
dbfr:averageSizePerMonth [ dc:date
"06/2002"^^xsd:gYearMonth ;
rdf:value "2610.66"^^xsd:float ]
;
(...)
dbfr:size "159049"^^xsd:integer ;
dc:creator [ foaf:nick "Rinaldum" ] ;
sioc:note "wikification"^^xsd:string ;
prov:wasRevisionOf <http:// … 119074391> ;
prov:wasAttributedTo [ foaf:name "Rémih" ; a prov:Person,
foaf:Person ] .
<http:// … 119074391> a prov:Revision ;
dc:created "2015-09-29T19:35:34"^^xsd:dateTime ;
dbfr:size "159034"^^xsd:integer ;
dbfr:sizeNewDifference "-5"^^xsd:integer ;
sioc:note "/*Années théâtre*/ neutralisation"^^xsd:string ;
prov:wasAttributedTo [ foaf:name "Thouny" ; a prov:Person,
foaf:Person ] ;
prov:wasRevisionOf <http://... 118903583> .
(...)
<http:// … oldid=118201419> a prov:Revision ;
prov:wasAttributedTo [ foaf:name "OrlodrimBot" ; a
prov:SoftwareAgent ] ;
(...)
[Gandon, Boyer, Corby, Monnin 2016]
24. Dataset description No. RDF triples
dataset description + definition of a few properties 170
articles metadata (title, authors, DOIs, journal etc.) 3 722 381
named entities identified by Entity-fishing in articles titles/abstracts 35 049 832
named entities identified by Entity-fishing in articles bodies 1 156 611 321
named entities identified by Bioportal Annotator in articles titles/abstracts 104 430 547
named entities identified by DBpedia Spotlight in articles titles/abstracts 65 359 664
argumentative components and PICO elements by ACTA from articles titles/abstracts 7 469 234
Total 1 361 451 364
COVID ON THE WEB
[Corby, Michel, Gazzotti, Winckler, et al. 2019]
25. CRAWLING
▪ Predict data availability
▪ Select features of URIs
▪ Learn crawling selection
(KNN/NaiveBayes/SVM)
▪ Online learning w. crawling
(FTRL-proximal algorithm)
[Huang, Gandon 2019]
26. QUERY
• automatically suggest relevant data sources to solve a query
• sets of path features: star, sink, chain
• approximate containment search: locality sensitive hashing
[Huang, Gandon 2020]
27. IndeGx: knowledge acquisition about knowledge graphs
declarative framework with extractions rules managed on git to build a public RDF index of
RDF datasets https://github.com/Wimmics/dekalog
[Maillot et al. 2022]
28. KartoGraphI: visualize what can be accessed through
the endpoint of IndeGx
23/06/2022
-
28
[Maillot et al. 2022]
Locations of datasets
http://prod-dekalog.inria.fr/
30. WEB EDGE ACQUISITION
▪ Edge AI directly in the browser
▪ Web APIs, models, protocols,…
[WebML @ W3C]
31. e.g. what types of knowledge have we missed?
What are the big (competency) questions of our time?
32. ALOOF: robots learning by reading on the Web
▪ First Object Relation Knowledge Base: 46212 co-mentions, 49 tools, 14 rooms, 101
“possible location” relations, 696 tuples <entity, relation, frame>
▪ Evaluation: 100 domestic instruments, 20 rooms, 2000 crowdsourcing judgements
▪ Shared between robots through a shared Web knowledge base
Annie cuts the bread in the kitchen with her knife dbp:Knife aloof:Location dbp:Kitchen
[Cabrio, Basile
et al. 2017]
33. ALOOF: RDF dataset about objects
[Cabrio, Basile et al.]
▪ common sense knowledge about objects: classification, prototypical locations
and actions
▪ knowledge extracted from natural language parsing, crowdsourcing,
distributional semantics, keyword linking, ...
35. “from point to point”, a book by
▪ semiotic dimension in knowledge models is both rare and needed
▪ semiotic knowledge acquisition is important for intelligent HCI
36. e.g. where does this knowledge come from?
What are the big (competency) questions of our time?
37. gaining knowledge from the Web influenza in the US
● Google ● CDC
http://tylervigen.com/spurious-correlations
limits
bias
38. is there a streetlight effect in our community?
should we acquire other types of (meta-)knowledge ?
48. PREDICT HOSPITALIZATION
▪ Physician’s records classification
to predict hospitalization
[Gazzotti, Faron et al. 2017]
Sexe Date Cause CISP2 ... History Observations
H 25/04/2012 vaccin-antitétanique A44 ... Appendicite EN CP - Bon état général -
auscult pulm libre; bdc
rég sans souffle - tympans
ok-
Element Number
Patients
Consultations
Past medical history
Biometric data
Semiotics
Diagnosis
Row of prescribed drugs
Symptoms
Health care procedures
Additional examination
Paramedical prescription
Observations/notes
55 823
364 684
187 290
293 908
250 669
117 442
847 422
23 488
11 850
871 590
17 222
56 143
49. PREDICT HOSPITALIZATION
▪ Physician’s records classification
to predict hospitalization
▪ Study enrichment and features
impact and combinations on
different ML methods
[Gazzotti, Faron et al. 2017]
Sexe Date Cause CISP2 ... History Observations
H 25/04/2012 vaccin-antitétanique A44 ... Appendicite EN CP - Bon état général -
auscult pulm libre; bdc
rég sans souffle - tympans
ok-
Element Number
Patients
Consultations
Past medical history
Biometric data
Semiotics
Diagnosis
Row of prescribed drugs
Symptoms
Health care procedures
Additional examination
Paramedical prescription
Observations/notes
55 823
364 684
187 290
293 908
250 669
117 442
847 422
23 488
11 850
871 590
17 222
56 143
(1)
(2)
50. RDFMiner : an evolutionary approach for OWL 2
axiom extraction
• Grammar-based genetic programming for mining subsumption axioms involving
complex class expressions
• Evolutionary axiom discovery from knowledge graphs based on “grammatical
evolution” and possibility theory.
[Félin, Tettamanzi, 2021]
52. (1)
Image Metadata Score
portrait
50350012455
C:Jocondejoconde0138m503501_d0012455-000_p.jpg
cheval:
0.999
Image Metadata Score
figure (saint Eloi de Noyon, évêque, en pied, bénédiction,
vêtement liturgique, mitre, attribut, cheval, marteau, outil :
ferronnerie)
000SC022652
C:/Joconde/joconde0355/m079806_bsa0030101_p.jpg
cheval:
0.006
MonaLIA
▪ reason & query on RDF to build training sets.
▪ transfer learning & CNN classifiers on targeted
categories (topics, techniques, etc.)
▪ reason & query RDF of results to address
silence, noise and explain
350 000 images
of artworks
RDF metadata based
on external thesauri
Joconde database from French museums
(3)
[Bobasheva, Gandon, Precioso, 2021]
(2)
53. Oedeepus complex…
▪ Is the knowledge acquired by neural networks really acquired for us?
▪ Latent knowledge of neural network is, by definition, not knowledge for us
▪ Explainable AI → Actionable Knowledge → Decision Making
54. User Scored Evaluation of Non-Unique Explanations for Relational Graph
Convolutional Network Link Prediction on Knowledge Graphs
▪ Traces of inferences as a ground truth for explanation
Rules equivalent to Horn clauses ¬𝑙1 ∨ ... ∨ ¬𝑙𝑛 ∨ 𝑙𝑐 or 𝑙𝑐 ← 𝑙1 ∧ ... ∧ 𝑙𝑛
e.g. ℎ𝑎𝑠𝐺𝑟𝑎𝑛𝑑𝑝𝑎𝑟𝑒𝑛𝑡 (𝑋, 𝑍) ←ℎ𝑎𝑠𝑃𝑎𝑟𝑒𝑛𝑡 (𝑋,𝑌) ∧ ℎ𝑎𝑠𝑃𝑎𝑟𝑒𝑛𝑡 (𝑌, 𝑍)
▪ Use the open source rule engine CORESE [4] to create a dataset with explanations using rules:
FrenchRoyalty-200k
Extraction and cleaning of family relationships of kings and queens of France
6 types of relations : hasBrother, hasSister, hasChild, hasParent, hasSpouse, hasGrandparent
▪ User evaluation and scoring of the rules/types of explanations
▪ Quantitative analysis and comparison of RGCN explanation methods for link prediction
▪ Qualitative analysis of the errors (wrong predicates, wrong subject/object,…)
+
Louis_VI_of_France
Agnes_of_France
Louis_VII_of_France
Adela_of_Champagne
Hugh,_Count_of_Champagne
Constance_of_France
hasChild
Bertha_of_Holland
hasParent
hasGrandparent
hasSpouse
hasGrandparent
hasChild
hasChild
hasParent
[Halliwell, Gandon, Lecue, 2021/2]
55. Entities-as-Experts model from Févry et al. (2020).
Facts as Experts: Adaptable and Interpretable Neural Memory over
Symbolic Knowledge Pat Verga, Haitian Sun, Livio Baldini Soares, William W. Cohen Google Research
Model including an explicit interface between symbolically interpretable factual memory and
subsymbolic neural knowledge - i.e. the new task includes the act of querying an explicit memory
○ include databases in addition to corpora to cover more facts / domains
○ update & augment the memory at inference time i.e. without retraining
preprint arXiv
56. e.g
• Knowledge Acquisition to provide, high quality, unbiased, documented, certified,
transparent, fair,… datasets/knowledge graphs/schemas/ontologies
• Approaches for knowledge-based auto-supervision
Knowledge is more than an input, an output or a parameter
Neural Network can be seen as form of (Knowledge) Graph
Acquire, represent, exchange meta knowledge of AI & enclosing decision workflow
eg. knowledge about NN as in FAIRNet ontology [Nguyen et al. arXiv:1907.11569]
deep knowledge acquisition
58. LOD Visualization Pipeline
Import
Retrieve data
with queries
Imported
dataset
Enriched
Dataset
Visual variables
(color, shape)
End User
Transform
Transform, enrich
and resample data
Visual Map
Map data to visual
representations
Rendering
Draw graphics
to display
Interaction
Selection, zoom,
etc
Web of Data
Data
visualization
[Menin, Winckler, et al. 2022]
60. Incremental Data Exploration
[Menin, Winckler, et al. 2022]
Follow-up Queries to dynamically import data into the visualization
during the exploratory process
65. PREDICT STUDENTS’ PATH
▪ a model of the students' learning
▪ predict success or failure to questions
▪ features from KG representations
▪ Logistic Regression (LR) / Factorization Machines
(FM) / Deep Factorization Machines (DeepFM)
[Rodriguez-Rocha, Faron, Ettorre, Michel et al. 2020]
Answers
Questions
s: students identifiers
q: questions identifiers
r: responses identifiers
a: number of attempts
w: number of wins
T: questions text embeddings
Q: graph embeddings of the questions
R: graph embeddings of the answers
e: extra group of calculated features:
question_difficulty,student_ability,
student_ability_progressive,
student_ability_progressive_question_difficulty
Features
66. almost 30 years ago… knowledge creation circle
symbol(ic AI) as the meeting level between natural and artificial intelligence
latent AI knowledge
is stuck here
the Web changed everything here
importance of explainable AI here
67. e.g. what are the access rights to a piece of knowledge?
What are the big (competency) questions of our time?
68. Context-aware Access Control for Linked Data
Shi3ld Access Control Manager
GET /data/resource HTTP/1.1
Host: example.org
Authorization: ...
SELECT …
WHERE {…}
http://wimmics.inria.fr/projects/shi3ld
69. meta-knowledge: access control to knowledge graphs
e.g. only my colleagues
working on the same subject
User
ASK{ ?res dcterms:creator ?prov .
?prov rel:hasColleague ?user .
?prov foaf:interestedBy ?topic .
?user foaf:interestedBy ?topic }
[Villata, Costabello, et al.]
73. learning a thesaurus from a folksonomy
flat folksonomies
web 2.0
pollution
soil pollution
has narrower
pollutant energy
related related
thesaurus
SKOS
[Limpens, et al. 2008-2011]
74. typed sociograms & parameterized metrics
Fabien
creator
author
Man
type
doc.html
author
Semantic web is not antisocial
Person
Man
sub property sub class
semantic web
title
Fabien
Marco Guillaume
Nicolas
Michel
Rémi
social network analysis
)
,
(
;
)
( p
x
rel
x
p
din =
4
)
( =
Guillaume
din
creator
Person
type
[Erétéo, et al. 2008-2011]
SNA indices SPARQL formal definition
)
(G
nbactor
type
select merge count(?x) as ?nbactor from <G>
where{
?x rdf:type param[type]
}
)
(G
nbactor
rel
select merge count(?x) as ?nbactors from <G>
where{
{?x param[rel] ?y}
UNION{?y param[rel] ?x}
}
)
(G
nbsubject
rel
select merge count(?x) as ?nbsubj from <G>
where{
?x param[rel] ?y
}
)
(G
nbobject
rel
select merge count(?y) as ?nbobj from <G>
where{
?x param[rel] ?y
}
)
(G
nbrelation
rel
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]) }
}
)
(G
Comp rel
select ?x ?y from <G> where {
?x param[rel] ?y
}group by any
)
(
, y
D dist
rel
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
)
(
, y
Din
dist
rel
select ?y count(?x) as ?indegree where{
?x (param[rel])*::$path ?y
filter(pathLength($path) <= param[dist])
}group by ?y
)
(
, y
Dout
dist
rel
select ?x count(?y) as ?outdegree where {
?x (param[rel])*::$path ?y
filter(pathLength($path) <= param[dist])
}group by ?x
)
,
( to
from
g rel
select ?from ?to $path pathLength($path) as
?length where{
?from sa (param[rel])*::$path ?to
}group by ?from ?to
)
(G
Diamrel
select pathLength($path) as ?length from <G>
where {
?y s (param[rel])*::$path ?to
}order by desc(?length)
limit 1
)
,
( to
from
nbg
rel
select ?from ?to count($path) as ?count
where{
?from sa (param[rel])*::$path ?to
}group by ?from ?to
75. • Formalized in RDF+OWL the activity traces (e.g. 4 years of interactions in StackOverflow)
• Topic Trees and Distributions (Latent Dirichlet Allocation) to detect overlapping communities
• Aligned to Linked Open Data and labeling detected communities with shared topic of interest
…
detecting overlapping epistemic communities and labeling
them with shared interest [Meng et al., 2013-2016]
76. e.g. what knowledge does the Web need?
What are the big (competency) questions of our time?
77. ARGUMENT MINING ON
CLINICAL TRIALS
▪ NLP, ML and arguments
▪ assist evidence-based medicine
▪ support doctors and clinicians
▪ identify doc. for certain disease
▪ analyze argumentative content
and PICO elements
[Mayer, Cabrio, Villata]
78. ARGUMENT MINING ON
POLITICAL SPEECHES
▪ NLP and Machine Learning.
▪ Support historians/social science scholars
▪ Analyze arguments in political speeches
▪ DISPUTool : 39 political debates,
last 50 years of US presidential
campaigns (1960-2016)
[Mayer, Cabrio, Villata]
80. The Web is much more than a social media application
81. What are these beasts from the 90s?
KACTUS
[Bernaras et al, 1996] MS & CQ
[Uschold, 1996]
SENSUS
[Swartout et al., 1997]
Methontology
[Fernández-López et al., 1997]
IDEF5
[Uschold, 1994]
What new methodology could we imagine leveraging all the facets of
the Web for all the methodological steps: collaboration space, corpora,
raw data, linked data, crowdsourcing, multilingual, multi-profiles,
mobile, …
82. A WEB IDENTIFYING EVERYTHING IRI & RESOURCES
identify on the networks anything we want
84. A WEB LINKING (META)DATA
RDF & LINKED DATA
distributed (meta)data networks
IRI & RESOURCES
identify on the networks anything we want
ex. the SOLID approach
85. A WEB LINKING DESCRIPTIONS
IRI
IRI
IRI
IRI IRI
IRI
IRI
86. A WEB LINKING COMPUTATIONS
RDF & LINKED DATA
distributed (meta)data networks
IRI & RESOURCES
identify on the networks anything we want
REST & LINKED COMPUTATIONS
decentralized computation networks
ex. Linked Functions in LDScript [Corby et al.]
ex. edge AI over the Web [WebML @ W3C]
ex. the SOLID approach
87. A WEB LINKING COMPUTATIONS
IRI
IRI
IRI
IRI IRI
IRI
IRI
f(x,y,z) {…}
88. A WEB LINKING ALL NETWORKS
RDF & LINKED DATA
distributed (meta)data networks
REST & LINKED COMPUTATIONS
decentralized computation networks
ex. Linked Functions in LDScript [Corby et al. 20xx]
ex. edge AI over the Web [WebML @ W3C]
ONTOLOGIES & LINKED SEMANTICS
extensible network of models
ex. FAIRNet ontology [Nguyen et al. arXiv:1907.11569]
IRI & RESOURCES
identify on the networks anything we want
ex. the SOLID approach
89. A WEB LINKING ALL NETWORKS
IRI
IRI
IRI
IRI IRI
IRI
IRI
IRI
IRI IRI
IRI
f(x,y,z) {…}
97. Terms & Conditions vs…
▪ ontological commitment vs introduction of bias vs informed consent
▪ an ontological commitment includes an informed consent
▪ tension : complexity of a commitment vs ergonomic interaction
32 241 words
18 301 words
15 352 words 36 275 words
19 972 words
11 195 words < < < < <
http://conversation.which.co.uk/technology/length-of-website-terms-and-conditions/
…Ontological Commitment
98. Knowledge-level Coupling is Complex
50 shades of commitment:
extend, same As, related to, complex alignment, …
99. “Knowledge is power” Sir Francis Bacon
“Code is law” Lawrence Lessig
“Knowledge encoding is a power law”me 😀
103. How to perform Sustainable/Frugal/Affordable
Knowledge Augmentation ?
• do we need EKFW 2023?
the knowledge forgetting conference?
• should we work on knowledge lossy compression,
aggregation, summary?
…
104. How to fight knowledge pollution?
• How to detect knowledge base spamming?
• How to avoid unknown / low knowledge quality?
unverified /unverifiable knowledge ?
• How to trace knowledge? (RDF1.2, RDF Hash…)
…
105. How to support knowledge augmentation?
• Should Knowledge acquisition be designed to be both ways?
For the machine and for us ?
• Can we reconcile user-centric and reusable knowledge engineering?
• Shouldn’t explainable knowledge acquisition be the first step to support
knowledge augmentation ?
• How to avoid building more distorting digital mirrors?
• How to detect knowledge bubbles and burst them?
…