Short presentation from a working group at the 2008 social web communities workshop held in September 2008 at the Dagstuhl in Saarbrucken. The presentation discusses the social aspects of the kinds of tools that could be built once a connected web of data was easily mined.
1. Closing Triangles at the Café Symantique
Harith Alani Ian Mulvany
Alexandre Passant
Alexander Löser
Christian Bizer
Peter Mika
Nicolas
Maisonneauve Ciro Cattuto Christian Bauckhage
topics of the talk:
- connecting data sources
- connecting the real world
These connections can be considered as closing triangles across hyper-dimensional networks
Key issues raised include:provenance, accurate profiling, disambiguation, privacy, pushing
and polling data
Will discuss
- a real world example
- how to do this
- what does it mean, and what does it give us in our lives
2. The talk takes a global view
we assumed that all of the nitty gritty problems would be solved
(we recognize many of the problems and believe them to be tractable)
Wanted to take a more discursive approach.
3. Let‘s assume you are hungry and
you look for a restaurant
We wanted to look at a real world scenario to ground our thinking and we settled on this question,
4. In 1974 you would
• Call 2 friends for recommendations ($0,40)
• You only reach the one that has no idea
• Ask a taxi driver
• he recommends you a fast-food place
• Stroll through the street
… and probably reach the following restaurant
Process steps are:
Gathering The data
Trusting the data
Disambiguating
Understanding and analysing the data
closing triangles
5. Café Symantique
You find yourself, perhaps in an unfamilliar setting,
The question is, do you go in to the Cafe ?
6. • What could we improve with 21th century
technology?
7. Google has answered only some of this for us
Finding some places is now easy, but how can we help with the decision on whether we should enter this place?
These recommendations donʼt show you how to find unpopular places, they appear off of the front page.
8. Whats the process?
• Gathering The data
• Trusting the data
• Integration / Disambiguating
• Understanding and analyzing the data
• closing triangles
In 1974 the process of gathering data is easy, but the data is poor,
Now merging the data is hard, but the potential for the data quality is high
9. Gatherin Trustin Integrat Analyz Triangl
g g ing ing es
del.icio.us
An issue with merging data is that the data exists across many different islands
- rfid, fire eagle point the way to merging these islands with the real world
- we assume that these data sources can be combined
10. Gatherin Trustin Integrat Analyz Triangl
g g ing ing es
Trust ?
what do you do when you have 34k friends?
can we convince people to trust collaborative filters more than their friends?
11. Gatherin Trustin Integrat Analyz Triangl
g g ing ing es
Privacy
• Social graph fragmentation / delivering
issues
• Deciding which data you will deliver to
whom
• oAuth / OpenID / Social networking
policies
Want to ensure that when we merge data we merge the correct personas
12. Analyz Triangl
Gather Trust Integrate
e es
• Tag cloud merging
– Disambiguation
– Individual/Community tag frequency
– Tag Concept
– Syntactical analysis
• Building profiles of interest
How do we understand mixed signals from different sources?
13. Gatherin Trustin Integrat Analyz Triangl
g g ing ing es
Itʼs clear that tags taken from more than one source will give us a stronger sense of
the ground truth of the personomy of a person
14. Gatherin Trustin Integrat Analyz Triangl
g g ing ing es
Rated 5/5 Rated 1/5
Redemption Based-on-Play
Android Love Refugee
Spacecraft
Time-Travel Soldier Famous-Score Hope
Alien
Blockbuster Alien Broken-Heart Blockbuster
Space
War
Futuristic Based-on-Novel Racism
Artificial-Intelligence Hero Melodrama
15. Gatherin Trustin Integrat Analyz Triangl
g g ing ing es
Can use semantic tools to help with disambiguation
16. • But does this tool make you happy?
However an important question to ask
17. C’mon, Be Happy
• Hope (… find the secret little grandma style
restaurant)
• Belonging ( … to the small insider group
knowing the secret restaurant)
• self esteem (be the first one found it …)
• more more, optimization (it took you only 30
minutes … )
• Security (gov reports mean you know the place
won’t poison you )
Look to marketing to tell us what the drivers of happiness are
18. • Not just about friending people
• Connect people to places
• Connect people to things
In our discussions we felt strongly that the web of data is about
connecting more than just people to people, that novel, surprising
and fun tools could be built on top of the frameworks described at this meeting.
19. Can we connect a place that you are walking along with a book that you have read?
Can you be presented with a piece of music at a location that a friend of yours listened to at
some point in the past at that same location?
This is a mix between serendipity and reality mining
20. Can we connect a place that you are walking along with a book that you have read?
Can you be presented with a piece of music at a location that a friend of yours listened to at
some point in the past at that same location?
This is a mix between serendipity and reality mining
21. Can we connect a place that you are walking along with a book that you have read?
Can you be presented with a piece of music at a location that a friend of yours listened to at
some point in the past at that same location?
This is a mix between serendipity and reality mining
22. • Adds to the delight in our lives
• More Happy, make numinous
Can we connect a place that you are walking along with a book that you have read?
Can you be presented with a piece of music at a location that a friend of yours listened to at
some point in the past at that same location?
This is a mix between serendipity and reality mining
23. How do we map „happy“ as a
multi-dimensional-vector?
• V = {?,? …. ?}
• where ? in {who, what, where, when, why}
two key challenges to this community
- define the vector of happy
cost functions are defined against an assumed need, our needs in this context are not so well defined as we wish to accentuate the element of surprise
in the lives of people
- easily tie interrogative attributes to triples, or what have you, by context such as person, event, location, time or reason