Federating advertisement
targeting with Linked Data
Sven Lieber, Ben De Meester, Ruben Verborgh and Anastasia Dimou

Sven Lieber, Ben De Meester, Ruben Verborgh and Anastasia Dimou
Federating advertisement
targeting with Linked Data

An online advertising example
Federated querying with EcoDaLo
Comparison of different approaches
Privacy and ethics considerations
3

Although still relying on
an identification mechanism,
we improve advertising targeting
with decentralized knowledge graphs
by reusing existing infrastructure
which avoids data sharing.

An online advertising example
Federated querying with EcoDaLo
Comparison of different approaches
Privacy and ethics considerations
5

How does online advertising work?
C
Content
what: promote SEMANTICS conference
target: SemWeb researchers between 18 and 60
format: mobile leaderboard
Online
Behavioral
Advertising
(OBA)
Campaign
setup
Ad
serving

An online advertising example
Publisher A
Visitor traits
Publisher A
Observations
Ad
Server
E.g.
“between_18_and_25”
C
User visiting a website
with a mobile device Publisher B
Visitor traits Publisher BObservationsE.g. “LD_Expert”
Combined
data
Data integration
what: promote SEMANTICS conference
target: SemWeb researchers between 18 and 60
format: mobile leaderboard
X Privacy considerations
X Competitive disadvantage

An online advertising example
Federated querying with EcoDaLo
Comparison of different approaches
Privacy and ethics considerations
8

EcoDaLo facts
Research in Flanders, Belgium
Three complementary funding consortium partners:
AdLogix, Pebble Media and Roularta Media Group
Partners explaining our solution
https://sven-lieber.org/ecodalo-video

An online advertising example
Publisher A
Visitor traits
Publisher A
Observations
Ad
Server
E.g.
“between_18_and_25”
C
User visiting a website
with a mobile device Publisher B
Visitor traits Publisher BObservationsE.g. “LD_Expert”
Combined
data
Data integration
what: promote SEMANTICS conference
target: SemWeb researchers between 18 and 60
format: mobile leaderboard
X Privacy considerations
X Competitive disadvantage

Our approach avoids combination of data
Publisher A
Visitor traits
Publisher A
Observations
Ad
Server
E.g.
“between_18_and_25”
C
User visiting a website
with a mobile device Publisher B
Visitor traits Publisher BObservationsE.g. “LD_Expert”
what: promote SEMANTICS conference
target: SemWeb researchers between 18 and 60
format: mobile leaderboard
Federated
Querying
Only aggregate traits
No data is shared
Exclusivity of traits (and how observed)

Our solution is based on declarative semantic mappings
Publisher A
Visitor traits
E.g.
“between_18_and_25”
Publisher B
Visitor traits
E.g. “LD_Expert”
Common
Trait
Model
(SKOS)
Ad
Server
Federated
Querying
Between_18_and_60 includes between_18_and_25
Taxonomic relations between SemWeb_Expert and
LinkedData_Expert or Ontology_Expert
C

An online advertising example
Federated querying with EcoDaLo
Comparison of different approaches
Privacy and ethics considerations
13

Our solution compared to other approaches
Local publisher Global publisher (Data)
integration
(Trait) federation
Trait quality
Scale
Exclusive(privacy)
Ease of setup
Interoperability
Maintainability
X
XX XX
X
X
X
X
X
XXX

An online advertising example
Federated querying with EcoDaLo
Comparison of different approaches
Privacy and ethics considerations
15

GDPR-compliant consent needed
Which personal data is used for which purpose; third parties; explicit opt-in
Ethical considerations beyond our technical solution
EcoDaLo assumes good faith of publishers, ethical guidelines exist which need
to be considered

An online advertising example
Federated querying with EcoDaLo
Comparison of different approaches
Privacy and ethics considerations
17

More research on describing and
using constraints for querying of
decentralized knowledge graphs
SvenLieber sven-lieber.orgknows.idlab.ugent.be