Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Unwinding The Twine
1. 4th Project Meeting - 15/12/2009 @ Munich
Unwinding the twine
a Linked Data approach to the user profiling
Davide Palmisano
Tuesday, December 15, 2009
2. User profiling and context models
a (very) short ToC
Collecting data from the Social Web [1]
a quick recap
Reasoning over them to build user profiles
with Linked Data
User profiles syndication with OpenSocial
and a quick demo
[1] http://bit.ly/82qqoX
Tuesday, December 15, 2009
3. User profiling and context models
user activity aggregation...
Tuesday, December 15, 2009
4. User profiling and context models
user activity aggregation...
What are such user graphs?
RDF named graphs
a set of RDF statements describing a user activities
on the Social Web
a complex twine acting as a uniform user log
just something a bit better than a collection of URLs
Tuesday, December 15, 2009
5. User profiling and context models
user activity aggregation...
But unfortunately,
a collection of URLs cannot be considered
a user profile
a user profile should represents interests,
behaviors, favorite music or movie genres,
preferred actors ...
how we can access to this information?
Tuesday, December 15, 2009
6. User profiling and context models
user activity aggregation...
The Beancounter[2], now has all the facilities needed
to interact with several identity resolvers
an identity resolver is a lightweight Web service[3]
able to return some Linked Data URIs given a
certain kind of URL or other type of identifiers
ISBN to URIs, Last.fm identifier to URIs ...
Tuesday, December 15, 2009
8. User profiling and context models
data enrichment
linking every URL to its LoD representative URIs allows
us to access to an incredible source of
information where infer user interests
SKOS subjects dbpedia:Category:American_film_actors
dbpedia:Category:American_film_directors
genres
dbpedia:Alternative_rock
related resources dbpedia:Harvard_University
Tuesday, December 15, 2009
10. User profiling and context models
so the idea is to collect such URIs, aggregate them for
each user registered to the Beancounter and represent
them as weighted foaf:interests
Tuesday, December 15, 2009
11. User profiling and context models
a real example
Salvatore is an enthusiast of Last.fm and is using it
everyday. He is registered to the Beancounter and he
gave his Last.fm credentials to it.
RAI wants to develop an application lettings Beancounter
users to get personalized news
how to achieve such scenario?
Tuesday, December 15, 2009
12. User profiling and context models
a real example
All the RAI developers need to do is:
completely delegate the user profiling mechanism
to the Beancounter,
persist the Beancounter returned user IDs,
access to the user profiles simple making OpenSocial
REST calls
Tuesday, December 15, 2009