This document summarizes research into improving experience sharing through social networks. A survey found that people primarily learn about events from friends but want better tools to discover new activities. An experiment compared restaurant recommendations from friends on Facebook to ratings from the general public. Creating recommendation networks based on friendship and similarity improved precision and recall over general ratings. The researchers plan to conduct a user study and expand experiments internationally to further develop their experience sharing tools.
1 of 28
Download to read offline
More Related Content
Comealong: Empowering Experience-Sharing through Social Networks
1. Empowering Experience-Sharing through Social Networks
Beatrice Valeri, Marcos Baez, Fabio Casati
University of Trento, Italy
{valeri, baez, casati}@disi.unitn.it
Tuesday, October 16, 12
5. Experience-sharing
Do current systems really
empower our experiences?
Tuesday, October 16, 12
6. Experience-sharing
鍖nding activities you really like
Do current systems really
empower our experiences?
Tuesday, October 16, 12
7. Experience-sharing
鍖nding activities you really like with the right people
Do current systems really
empower our experiences?
Tuesday, October 16, 12
8. Experience-sharing
鍖nding activities you really like with the right people without that 鍖lling that your missing out things
Do current systems really
empower our experiences?
Tuesday, October 16, 12
9. Survey
?
?? Sources of event info
Aspects considered
when deciding
Ability to 鍖nd
interesting events
Tuesday, October 16, 12
10. Survey
?
?? Sources of event info
Aspects considered
when deciding
Ability to 鍖nd
interesting events
~150 participants
Tuesday, October 16, 12
11. RESULTS Survey
Sources of event info
Invitation / Information from
Friends
Aspects considered
when deciding
44% often look at who goes
to decide
Tuesday, October 16, 12
12. RESULTS Survey
Ability to 鍖nd
interesting events
Id like to have some way to
鍖nd interesting things to do
Sources vs ability to
鍖nd events
People using the web feel
the need for better tools
Tuesday, October 16, 12
13. Experiment
75 restaurants around
Trentino - top TA
rating 1-5 from local
people
friends from Facebook
Tuesday, October 16, 12
18. Comparing to TA
The population has a
signi鍖cant impact
Tuesday, October 16, 12
19. A deeper look
De鍖ned 4 different
networks
- overall
- friends
- similar
- similar friends
Evaluated how effective
they are to recommend
Tuesday, October 16, 12
20. Metrics
A/B testing (70-30) per
user
Precision: + 鍖nding out
something interesting
to do
Recall: + reduce feeling
of missing out things
Tuesday, October 16, 12
21. Results
Similarity and friendship
show promising results
features for similarity
should be explored
Tuesday, October 16, 12
27. Whats next
Run a user-study with
initial subjects
Run experiments in
other countries
(Paraguay, Russia)
Pilot involving local
businesses
Tuesday, October 16, 12