8. Cos’è
una
rete
sociale?
Traditional Media
Broadcast Media: One-to-Many
One to Many
Communication Media: One-to-One
9. Characteristics na
Socialsociale?
Cos’è
u of rete
Media
Everyone can be a media outlet
• User
generated
content
Disappearing of communications barrier
• Rich Usernriched
content
User
e Interaction
User-Generated Contents
User
interacCon
• User Enriched Contents
• User developed widgetsubiqua
Comunicazione
Collaborative environment
• Collective Wisdom environment
CollaboraCve
C
Long Tail
• …
Broadcast Media Social Media
Filter, then Publish Publish, then Filter
14. I
websites
più
visitaC
• Il
traffico
internet
più
alto
(daC
Alexa,
oQobre
2012)
1
Facebook
11
Blogspot
2
Google
12
LinkedIn
3
YouTube
13
Taobao
4
Yahoo
14
Google
India
5
Baidoo
15
Yahoo
Japan
6
wikipedia
16
Sina.com.cn
7
Windows
live
17
msn
8
TwiQer
18
Google
hk
9
QQ.com
19
Google
de
10
Amazon
20
Bing
19. L’azienda
e
i
social
networks
• Pubbliche
relazioni
• Customer
Support
• Market
Research
• Brand
MarkeCng
• PromoCons
• Consumer
EducaCon
• Sales
• New
Product
Development
• Customer
RelaConship
Management
21. DaC,
daC,
daC
• Grossi
volumi,
grande
varietà
– Milioni
di
utenC,
milioni
di
contenuC
– testuale,
MulCmediale
(immagini,
video,
etc.)
– Milioni
di
connessioni
– Tendenze,
preferenze,
comportamenC,
…
• I
daC
sono
open
e
facili
da
accedere
– Facili
da
reperire
– Di
pubblico
dominio
–
Developers
APIs
–
Spidering
the
Web
22. Le
opportunità
Any user can share and contribute
content, tente
può
condividere
e
• Ogni
u express opinions, link to
others
contribuire
ai
contenuC,
This means: Can data-mine opinions
esprimere
opinioni,
collegarsi
ad
and behaviors of millions of users to
altri
gain insights into:
• Questo
significa:
Human behavior
– Human
behavior
Marketing analytics
– MarkeCng
analyCcs
Product sentiment
– Product
senCment
8/21/2011 Jure Leskovec:Social Media Analytics (KDD '11 tutorial) 6
23. Actionable
Intelligence
Consumer Generated,
Not Edited,
Not Authenticated
8/21/2011 Jure Leskovec:Social Media Analytics (KDD '11 tutorial) 7
24. Applicazioni:
ReputaCon
management
• Consumer
Brand
AnalyCcs
– Cosa
dice
la
gente
sul
mio
marchio?
• MarkeCng
CommunicaCons
– Determinare
se
le
campagne
che
pianifico
saranno
efficaci
• Product
reviews
– Estrazione
automaCca
di
review
e
informazioni
su
prodom
e
servizi
• Facile
da
usare,
confortevole,
prezzo
adeguato,
…
25. Applicazioni:
Responsività
• CiCzen
response
• feedbacks
su
temaCche
poliCche
• Campagne
poliCche
– Perché
la
gente
supporta
un
candidato?
• Law
enforcement
– MovimenC
dissidenC
su
TwiQer
– Minority
report
hQp://www.nyCmes.com/2011/08/16/us/16police.html?_r=1
26. Applicazioni:
Viral
MarkeCng
• Viral
markeCng:
– Raccomandazioni
personlizzate
•
Il
ruolo
dei
forum
online:
– 79.2%
dei
partecipanC
ai
forum
aiutano
gli
utenC
connessi
a
prendere
decisioni
relaCve
a
un
prodoQo
– 65%
dei
partecipanC
ai
forum
condividono
consigli
(offline
o
personalizzaC)
basaC
sulle
informazioni
che
hanno
leQo
online
hQp://www.socialmediaexaminer.com/new-‐studies-‐show-‐value-‐of-‐social-‐me
27. Applicazioni:
Human
Behavior
analysis
Process social media content, provide tools
• for analysts to:ontenuC,
e
usufruire
di
tools
per
Processare
I
c
– IdenCficare
rnetworks: groups, members
Identify social eC
sociali:
gruppi,
membri
– IdenCficare
tand sentiment
Identify topics opics
e
senCments
Predictive
Modeling
Link Diagrams
Social Media
Content
8/21/2011 Jure Leskovec:Social Media Analytics (KDD '11 tutorial) 12
28. Relevance,
Authority,
SenCment
Page 27
• Le
tre
dimensioni
dell’interazione
sociale
ilities to pro-
ogs.
and market
s are a very
dia because Figure 1: Relevance, authority and sentiment at the blog level.
ly inaccessi- IBM’s
topic-‐based
blog
evaluator
c customer insights and opin- Finding the Relevant Blogs
an address several interesting OUR FIRST OBJECTIVE is to filter the vast blogosphere
29. SenCment
DetecCon
• E’
possibile
caraQerizzare
(in
maniera
automaCca)
il
tono
di
una
discussione?
ORMS3701_FTRs 2/3/10 4:56 PM Page 28
M A R K E T I N G & S O C I A L M E D I A
belief of the sentiment
IBM
Social
Media
AnalyCcs
associated with it. It is pos-
sible to learn from such
labeled words in conjunc-
tion with labeled docu-
ments. Furthermore, the
selection of words and doc-
uments to be labeled can be
made algorithmically.
Such an approach is
known as active dual super-
vision [5], and it can greatly
reduce the effort required to
label examples in a new
domain. Even though there
are expressions of sentiment
that are domain-specific,
Figure 2: Identifying and addressing negative sentiment.
there is still a large amount of
overlap in how positive and
30. Misurare
Influence
e
Authority
• Chi
sono
gli
utenC
suscembili?
• Come
si
propaga
un’informazione?
• Quando
un’opinione
è
affidabile?
31. Emerging
Topics
• Higher-‐level
concepts
dall’informazione
che
si
distribuisce
• Come
variano
quesC
concem?
hQp://memetracker.org
Most
menConed
phrases
in
the
US
presidenCal
campaign
32. I
social
media
e
le
imprese…
• Due
prospemve
– Nuovi
scenari
e
modelli
di
interazione
– AnalyCcs
• StreQa
cooperazione
con
ricerca
e
innovazione
– Nuovi
challenges
– Opportunità
enormi