Future TV is Now: Personalized & Social

Future TV is Now: Personalized & SocialLora Aroyo

VU University Amsterdam

@laroyoMonday, July 16, 12

Monday, July 16, 12

Monday, July 16, 12

Monday, July 16, 12

Monday, July 16, 12

Monday, July 16, 12

Monday, July 16, 12

Observations I

The TV world was closed, but now ...

• TV is the Web

• TV is Social

Monday, July 16, 12

Observations II

Personalized access to TV content is demanded, but ...

• data challenges multiply with Web openness• good social recommendations are not trivial • it is not about (single) algorithms any more• it is about metrics

Monday, July 16, 12

Observations III

Paradigm shift in personalized applications methodology

Monday, July 16, 12

Challenges

Monday, July 16, 12

Web vs. TV

TV is the Web

© Ugli. on the Walls of BBC London

Monday, July 16, 12

© Dan BrickleyMonday, July 16, 12

TV is SocialMonday, July 16, 12

Monday, July 16, 12

42% of UK adults who use Internet while watching TV also discuss or comment on programs they are watching

© Ericsson ComsumerLab: TV & Web Consumer Trends 2011

Monday, July 16, 12

Personalization Needed© Kierstin Shaylee

Monday, July 16, 12

choices, choices ...

© VickyBuser, BBC

Monday, July 16, 12

lost in space© RedBee Slides at MIPCube2012

Monday, July 16, 12

always missing something© RedBee Slides at MIPCube2012

Monday, July 16, 12

demand for experienceMonday, July 16, 12

Recommendations© Ericsson ComsumerLab: TV & Web Consumer Trends 2011

Monday, July 16, 12

Social Recommendations© Ericsson ComsumerLab: TV & Web Consumer Trends 2011

Monday, July 16, 12

Social Recommendations

I usually find NEW programmes to watch by:

RESULTS OF NOTUBE SOCIAL WEB & TV SURVEY 2012

© NoTube Studies @ BBC

Monday, July 16, 12

Social Recommendations

I usually find NEW programmes to watch by:

RESULTS OF NOTUBE SOCIAL WEB & TV SURVEY 2012

Social TV

RESULTS OF NOTUBE SOCIAL WEB & TV SURVEY 2012

© NoTube Studies @ BBC

Monday, July 16, 12

© RedBee Slides at MIPCube2012

Monday, July 16, 12

Social & Personalized TV

© Ugli. on the Walls of BBC London

Monday, July 16, 12

ScenariosMonday, July 16, 12

User Perspective

• what is the role of social recommendations?• how do people watch TV together?• how devices influence watching?• what are perceived trade-offs for privacy vs. personalization?

Monday, July 16, 12

Technology Perspective

• what is the role of metadata in TV & Apps?• what are ways to share information in real time?• what are ways to sync TV & other metadata?• what is an easy way for devices to find & talk to each other?

Monday, July 16, 12

http://notube.tvMonday, July 16, 12

http://www.youtube.com/watch?v=0QUPKQLvOHU

Monday, July 16, 12

the old TV model is broken

closed, proprietary, static, impersonal

Monday, July 16, 12

• access to basic metadata about programs • links are the basic currency of social media

• URIs for the things you watch • links to related Web entities

• even a small amount of metadata enables interesting apps

Open Web Standards & Data

Monday, July 16, 12

TV Program Enrichment

SemanticPattern-based

Recommendation Strategy

RDF GraphTV

Programs

Semantic ContentPatterns for

TV Programs

HybridRecommendation

Strategy

StatisticalSimilarity-based

Recommendation StrategyUser Ratings &

Demographics(BBC EPG

Data)

EPG Metadata(BBC)

Recommendation Service

SimilarityClusters

of Programs

User Data Analysis

End-UsersEnd Users

Linking Web Data to User ProfilesMonday, July 16, 12

TV Preference Data: Sparse

• Even for a single service (e.g. Netflix)

• NoTube’s open systems: challenges multiply

• often no global view, only per-user data

• many ways of identifying same content item

• many ways of identifying same user

• many ways of identifying entities e.g. actors, directors, ...

© Dan BrickleyMonday, July 16, 12

TV Preference Data: Fragmented

© Dan BrickleyMonday, July 16, 12

TV Preferences in NoTube

• Inferred from the Social Web

• tweets, FB likes, last.fm listened, etc.• weighted interests

• Represented using Linked Data web identifiers

• record-linkage:

• NLP:

facebook.com/pages/ dbpedia.org/

http://dbpedia.org/resource/Matt_Lucashttp://dbpedia.org/resource/David_Walliams

#littlebritain

Monday, July 16, 12

TV Preferences in NoTube

• Inferred from the Social Web

• tweets, FB likes, last.fm listened, etc.• weighted interests

• Represented using Linked Data web identifiers

• record-linkage:

• NLP:

facebook.com/pages/ dbpedia.org/

Brilliant british humor by Matt Lucas & David Walliams - whole range of facinating characters portraying diversity of british society

http://dbpedia.org/resource/Matt_Lucashttp://dbpedia.org/resource/David_Walliams

#littlebritain

Monday, July 16, 12

NoTube’s Beancounteraggregate, analyze & profile

Monday, July 16, 12

NoTube’s User Activitiesaggregate

Monday, July 16, 12

NoTube’s User Interestsprofile

Monday, July 16, 12

NoTube’s User Explanationscontrol

Monday, July 16, 12

Recommendations in NoTube

• surface interesting, new & relevant programs to individual and group users• combine in a complementary way different statistical & semantic approaches• define metrics for, e.g. serendipity, diversity, relevance

© Dan Brickley

Monday, July 16, 12

Analysis of LOD SourcesDataset LinkedMDB DBpedia Freebase

#triples 6,147,978 385,000,000 337,203,427

#props 221 1643 n.a.

#types 53 3,640,000 12,000,000

Properties #triples

foaf:page 512,944

http://purl.org/dc/terms/date 74,111

foaf:made 53,180

linkedmdb:actor_actorid 50,603

dbprop:hasPhotoCollection 30,354

Dataset Enriched BBC

#triples 240,630

#props 41

#types 17,029

Type Pattern #

People person - co_participation - person 62162

People person - actor - person 24843

Format program - has_format - format 13108

Genre program - has_genre - genre 12767

People person - influencedBy - person 10053

People person - partner - person 10053

Award program - award_won - award 782

Monday, July 16, 12

From Predictability to Serendipity

Monday, July 16, 12

From Predictability to Serendipity

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

Monday, July 16, 12

From Predictability to Serendipity

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

documentaryformat

format

format

Monday, July 16, 12

From Predictability to Serendipity

homes and gardens

genre

Grammy Awards

genre

genre art culture and the media

award

award

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

Basic patternsdocumentaryformat

format

format

Monday, July 16, 12

From Predictability to Serendipity

homes and gardens

genre

Grammy Awards

genre

genre art culture and the media

award

award

PaulMerton

anchor

partner

CarolineQuentin

Life ofRiley

actor

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

Basic patternsHomogeneous patterns

documentaryformat

format

format

comedy

genre

Monday, July 16, 12

From Predictability to Serendipity

homes and gardens

genre

Grammy Awards

genre

genre art culture and the media

award

award

PaulMerton

anchor

partner

CarolineQuentin

Life ofRiley

actor

SpikeMilligan

influence

The adventuresof Barry

McKenzie

actor

film

genr

e form

at

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

Basic patternsHomogeneous patterns

documentaryformat

format

format

comedy

genre

Monday, July 16, 12

From Predictability to Serendipity

homes and gardens

genre

Grammy Awards

genre

genre art culture and the media

award

award

PaulMerton

anchor

partner

CarolineQuentin

Life ofRiley

actor

SpikeMilligan

influence

The adventuresof Barry

McKenzie

actor

film

genr

e form

at

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

CharlieChaplin

synopsisenrichment

ShanghaiKnights

form

at

action andadventure

genre

JackieChan

actor

influence

Basic patternsHomogeneous patternsHeterogeneous patterns

documentaryformat

format

format

comedy

genre

Monday, July 16, 12

From Predictability to Serendipity

homes and gardens

genre

Grammy Awards

genre

genre art culture and the media

award

award

PaulMerton

anchor

partner

CarolineQuentin

Life ofRiley

actor

format

synopsisenrichment

Alfred Hitchcock

director

Suspicion

AcademyAwards

awar

d

Thriller

genre

SpikeMilligan

influence

The adventuresof Barry

McKenzie

actor

film

genr

e form

at

Paul Merton looks at

Alfred Hitchcock

SecretGardens

Make ‘em laugh

CharlieChaplin

synopsisenrichment

ShanghaiKnights

form

at

action andadventure

genre

JackieChan

actor

influence

Basic patternsHomogeneous patternsHeterogeneous patterns

documentaryformat

format

format

comedy

genre

Monday, July 16, 12

Weak Semantics & Strong Statistics

Monday, July 16, 12

Weak Semantics & Strong Statistics

Paul Merton

looks at Alfred Hitchcock

PaulMerton

anchor partnerCarolineQuentin

Life ofRiley

act

ordocumentary

form

at

art culture and the media

genreMonday, July 16, 12

Weak Semantics & Strong Statistics

Paul Merton

looks at Alfred Hitchcock

PaulMerton

anchor partnerCarolineQuentin

Life ofRiley

act

ordocumentary

form

at

art culture and the media

genre

woman25 years old1 child

Userprofile

Monday, July 16, 12

Weak Semantics & Strong Statistics

Paul Merton

looks at Alfred Hitchcock

PaulMerton

anchor partnerCarolineQuentin

Life ofRiley

act

ordocumentary

form

at

art culture and the media

genre

25 year old women with

1 child

woman25 years old1 child

Userprofile

Monday, July 16, 12

Weak Semantics & Strong Statistics

Paul Merton

looks at Alfred Hitchcock

PaulMerton

anchor partnerCarolineQuentin

Life ofRiley

act

ordocumentary

form

at

art culture and the media

genre

25 year old women with

1 child

Similar users

woman25 years old1 child

Userprofile

Monday, July 16, 12

NoTube’s N-Screen

N-Screendrag & drop, sharing & TV control

drag & drop, shuffle, sharing & TV control

Monday, July 16, 12

NoTube’s N-Screenexplore program

Monday, July 16, 12

NoTube’s N-Screensimilar programs

Monday, July 16, 12

• means of surfacing content buried in the video collection• when users might reach a dead-end with the recommendations • adds an extra element of serendipity

NoTube’s N-Screenrandom selection

Monday, July 16, 12

NoTube’s N-Screenexplanations

• depending on recommendation strategies• depending on the user’s task

Monday, July 16, 12

♥

INTERESTING PROGRAMMES

watching with friends

suggestions based on programmes others have watched

linking programmes by guests, actors, presenters etc.

sharing suggestions with others in the same room

?

watching 'together apart' in different locations

changing programmes on someone else's TV

♥♥

♥

✔

✘

suggestions based on your past 'Social Web' activities

controlling the TV

drag & drop

swapping suggestions with specific friends

sharing them

✔

♥

✔

♥

sharing to watch immediately

http://notube.tv

? explaining recommendations

'shuffle' option for programme selection

?

finding

USER TESTING RESULTS December 2011, BBC R&D London10 participants : 5 male, 5 female, aged between 20 to 64

drag & drop to change what's playing on TV

♥ ✔ ✘ ? loved liked disliked unsure

For more details:

Monday, July 16, 12

More from NoTube ...

http://notube.tv http://www.slideshare.net/NoTubeProject

@notubeproject

© Dan Brickley

Monday, July 16, 12

Challenges

• data challenges multiply with Web openness

• good social recommendations are preferred

• it is not about (single) algorithms any more

• it is about the metricsMonday, July 16, 12

What’s next

• The move to IP - opportunity to discover what people watch in greater breadth and depth• gather consumers’ viewing behavior & video streams• combine them with enhanced EPG as input for a holistic live-stream data mining analysis

• generate a high-quality linked open dataset (LOD) describing live TV programming• market research on viewing behavior

Monday, July 16, 12

Thank you

• for their slides & visuals:• Dan Brickley, Valentina Maccatrozzo, • Vicky Buser, Libby Miller, • Davide Palmisano, and the whole NoTube team

Monday, July 16, 12

Monday, July 16, 12