1. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Vision
about
Content
Exploita)on
Yiannis
Kompatsiaris
(CERTH-‐ITI)
Concerta8on
mee8ng,
Brussels,
25
June
2014
2. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
hEp://www.puzzlemarketer.com/digital-‐social-‐brands-‐in-‐60-‐seconds/
(Apr,
2012)
3. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Combining
various
types
of
content
(1/2)
• AV
content
(movies,
TV,
YouTube,
visual
ambient
sensors)
con8nues
to
be
a
major
source
of
informa8on
and
entertainment
• Social
Media
(UCG)
have
an
extremely
dynamic
nature
that
reflects
the
evolu8on
of
community
focus
(user’s
interests)
– When
aggregated
allows
the
detec8on
of
meaningful
topics,
events,
points
of
interest,
emo8onal
states
and
social
connec8ons
• Sensors
in
smartphones
and
wearable
devices
provide
personal,
real-‐8me
and
loca8on-‐based
user
feedback
• Addi8onal
sources
(e.g.
Open
Data)
4. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Combining
various
types
of
content
(2/2)
• In
most
exis8ng
R&D
approaches
and
applica8ons
each
type
is
considered
separately
or
with
limited
interac8on
• Analyze
each
content
type
in
order
to
reach
a
common
representa8on
level
(user,
seman8cs,
context
related)
and
then
correlate
the
various
sources
in
real-‐)me
• Present
the
result
in
an
effec8ve
and
engaging
way
for
a
variety
of
applica8ons
for
personal
consump)on
and
decision
making
5. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Example:
Capturing
urban
dynamics
through…
Analysis,
Correla)on,
Presenta)on
and
Visualiza)on
• Informa8on
in
real-‐8me
back
to
the
content
producers
(e.g.
News)
and
ci8zens
• Decision
Making
for
the
authori8es
• Linking
to
Open
Data
2
Social
Media
and
UGC
• Discover areas, POIs and events as “seen”
by aggregation of social users contribution
• Issues directly reported by citizens in mobile
apps
www.clusttour.gr
1 AV Content
• TV, radio, newspapers, YouTube videos
monitoring for local – city issues
lganalytics.mklab.iti.gr
3
Ambient
and
personal-‐wearable
sensors
• Extract “knowledge” about urban rhythms
and locations, patterns of activity
6. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Example:
Predic)on
of
elec)on
results
• Based
on
(i)
Social
media
(TwiEer
sen8ment
volume)
and
(ii)
polls
• Close-‐to-‐exit
poll
accuracy
can
be
achieved.
It
also
predicted
the
percentage
of
the
first
party
with
the
greatest
accuracy
• AV
content
analysis
and
correla8on
could
further
improve
the
results
• Extension
to
real-‐8me
monitoring
-‐
predic8on
• Reduced
cost
for
an
exis8ng
product
with
a
strong
demand
http://www.socialsensor.eu/news/133-sensing-social-media-to-predict-eu-elections
7. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Crawling
OSN-‐UGC
Sensors
AV
Content
CONTENT
Fusion
Near-‐duplicates
Text/AV
Indexing
INDEXING
Open
Data
SNA
Sen)ment
-‐
Influence
Trends
-‐
Topics
MINING
Model
Building
Concepts
Relevance
Diversity
Popularity
RETRIEVAL&RANKING
Veracity
Interac)on
Responsiveness
Aggrega)on
VISUALIZATION
Aesthe)cs
Correla)on
detec)on
Consump)on
Decision
Support
–
Recommenda)on,
…
Infrastructure
8. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Challenges
–
Content
(Mining)
• Mul)-‐modality:
e.g.
image
+
tags,
videos
• Heterogeneous
sensors
and
content:
web
(inc
social
media),
TV,
cell
phones,
wearable
sensors,
cameras
• Rich
social
context:
spa8o-‐temporal,
social
connec8ons,
rela8ons
and
social
graph
• Inconsistent
quality:
noise,
spam,
ambiguity,
fake,
propaganda,
complementary
&
contradictory
content
• Huge
volume:
Massively
produced
and
disseminated
• Dynamic:
Fast
updates,
real-‐8me
• Mul)-‐source:
may
be
generated
by
different
applica8ons
and
user
communi8es
and
connected
to
other
sources
(e.g.
LOD)
9. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Policy
–
Licensing
–
Legal
challenges
(Social
Media)
•
Fragmented
access
to
data
– Separate
wrappers/APIs
for
each
source
(TwiEer,
Facebook,
etc.)
– Different
data
collec8on/crawling
policies
•
Limita8ons
imposed
by
API
providers
(“Walled
Gardens”)
• Full
access
to
data
impossible
or
extremely
expensive
(e.g.
see
data
licensing
plans
for
GNIP
and
DataSij
• Non-‐transparent
data
access
prac8ces
(e.g.
access
is
provided
to
an
organiza8on/person
if
they
have
a
contact
in
TwiEer)
•
Constant
change
of
model
and
ToS
of
social
APIs
– No
backwards
compa8bility,
addi8onal
development
costs
•
Ephemeral
nature
of
content
• Social
search
results
ojen
lead
to
removed
content
à
inconsistent
and
unreliable
referencing
•
User
Privacy
&
Purpose
of
use
• Fuzzy
regulatory
framework
regarding
mining
user-‐contributed
data
10. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Vision
–
future
direc)ons
(1/3)
Policies
-‐
Infrastructure
• Data
accessibility
– EC
can
work
towards
opening
of
“walled
gardens”
and
more
open
data
policies
– Standards
for
sensor/wearable
devices
data
access
– Technology
or
other
innova8ve
methods
be
used
to
make
data
more
accessible
to
society
at
large
• Privacy,
security
and
value
of
personal
data
• Infrastructure
– Large-‐scale
and
real-‐8me
data
analy8cs
create
requirements
for
cloud
and
common
EU
infrastructure
– E.g.
EC
organiza8on
providing
access
to
social
media
data
11. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Vision
–
future
direc)ons
(2/3)
Research
and
Development
• Indexing,
Mining
and
Retrieval
– Real-‐8me
and
scalable
approaches
• Efficiency
of
seman8cs
and
analysis
vs.
performance
vs.
infrastructure
– Fusion
of
various
modali8es
• Content,
social,
temporal,
loca8on
• Mul8modal
search
and
clustering
– Verifica8on
&
Linking
to
other
sources
(web,
Linked
Open
Data)
– Content
correla8on
and
cross-‐valida8on
• Contradictory,
complementary
and
duplicate
content
• Unreliable,
fake
content
• Visualiza8on
and
User
Interac8on
– Data
visualisa8on
and
storytelling
can
be
used
to
increase
our
understanding
of
data
• Beyond
“direct”
personaliza8on
– Mobile
Devices
and
applica8ons
– Augmented
Reality
can
be
used
to
mix
the
real
with
the
digital
world
and
combine
live
with
legacy
data
12. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Vision
–
future
direc)ons
(3/3)
Consump)on
-‐
Applica)ons
• Decision
support
– Inference/Predic)on
• in
real-‐8me,
beyond
current
and
past
insights
– Visualisa8on-‐assisted
– Summarisa8on
• New
applica8ons
– Large
scale
mul8modal
analysis,
retrieval
and
interac8on
in
• Smart
Ci)es
and
well
being
(linking
with
Open
Data)
• Environmental
monitoring,
forensics
• Health
data,
epidemics
and
disaster
monitoring
• Impact
–
Innova8on
– Business
models
and
strategies
for
profitable
commercial
services
13. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
Summary
of
future
direc)ons
• Develop
approaches
that
take
into
account
all
types
of
available
content
– Efficient
analy8cs,
linking
and
correla8on
• Scalability,
mul8-‐modality,
fusion,
visualiza8on,
user
interac8on
across
devices
• Build
applica8ons
in
domains
that
provide
value
for
users
and
decision
makers
• Address
policy
issues
14. Concerta)on
mee)ng,
Brussels,
25
June
2014
Y.
Kompatsiaris
CERTH-‐ITI,
SocialSensor
European
Centre
for
Social
Media
• Topics
–
Social
media
analy8cs
–
Verifica8on
–
Visualisa8on
–
Applica8ons
in
different
domains
• Ac8vi8es
–
Lis8ngs
of
project,
results,
ins8tu8ons,
events
–
Community
building
–
Support/organise
events
–
Common
social
media
presence
(e.g.
LinkedIn)
–
Funding
from
subscrip8ons,
training,
commercialisa8on
– Suppor8ng
projects:
SocialSensor,
Reveal,
MULTISENSOR,
PHEME,
DecarboNet,
MWCC,
uComp,
– Website:
hEp://www.socialmediacentre.eu/
– Research-‐academic:
STCSN
hEp://stcsn.ieee.net/
15. Thank
you
for
your
aEen8on!
ikom@i8.gr
hEp://mklab.i8.gr