3. 3
Participative systems
A particular class of social systems in which people can interact, share
information, or both.
Encyclopedias Maps Bookmarks
Music News Word processors
Politecnico di Milano, 07-04-2009 Davide Eynard
5. 5
Motivations
Successful
See “What is Web2.0” by Tim O'Reilly
On the Internet, but also inside intranets
Interesting from different point of views:
Psychology
• Incentives, bootstrap problem
HCI
• New interfaces and interaction paradigms
Social Sciences
• Collective intelligence, trust
KM
• Use meaningful formats for interoperability
New technologies
• Scalability, reliability
Politecnico di Milano, 07-04-2009 Davide Eynard
6. 6
There's Bad and “Bad”
Bad collaborative system
Too complex
Bad interaction with user
No incentives to participate
Tool doesn't fit the objective
“I love the system – I just don't use it!”
“Bad” collaborative system
Users like it
Fits tool and community well, but...
... it could be made better with semantics!
Politecnico di Milano, 07-04-2009 Davide Eynard
7. 7
Research objectives
Given a community, a task and a context
develop a methodology to evaluate whether a system
correctly fits an activity
use semantics to make the system better, incentivating
user contribution
Politecnico di Milano, 07-04-2009 Davide Eynard
9. 9
Background – Social Systems
Creating and publishing
• blogs, wikis, collaborative editors
Communicating
• e-mail, forums, chat, microblogging
Sharing
• p2p, client-server, social bookmarking
Recommending
• implicitly or explicitly, specific or general
Coordinating
• calendars, project or knowledge management
Networking
• social networks (object centered or not)
Playing
• MUD, MMORPG (Second Life)
Market places
• auctions, recruitment
Politecnico di Milano, 07-04-2009 Davide Eynard
10. 10
Background – Basics on social interactions
Wenger, Lave
• Communities of Practice
• Legitimate Peripheral Participation
• see Bryant, Forte, Bruckman:
Becoming Wikipedian
Image courtesy of Ross Mayfield's Weblog
Engeström, Vygotsky
• Activity Theory
Image courtesy of University of Helsinki - Center for Activity Theory
and Developmental Work Research
Politecnico di Milano, 07-04-2009 Davide Eynard
11. 11
Background – Semantic Web
“An extension of the current Web, in which information is given well
defined meaning, better enabling computers and people to work in
cooperation”
Politecnico di Milano, 07-04-2009 Davide Eynard
12. 12
Ch 3
Our work
Politecnico di Milano, 07-04-2009 Davide Eynard
13. 13
Our work – Evaluate and design
The Six Ws: User-centered design:
Who keep semantics hidden
• user, community, producer, reuse and standardize
consumer merge with the activity
What
evolve with complexity
• tool, object, contents, traces
How
(see D. Norman: “The design of
• explicit vs implicit
everyday things”)
participation, coupling
When
• sync vs async participation
Where
• context, centralized vs
distributed
Why
• incentives, engagement
Politecnico di Milano, 07-04-2009 Davide Eynard
14. 14
Our work – Extend systems with semantics
Semantics have been applied on different levels
Link data (“a little semantics goes a long way”, J. Hendler)
Better describe knowledge
Infer new knowledge with reasoning
When ontologies are used, they are applied on different levels too
Upper
Domain
Context
System [B.1]
Politecnico di Milano, 07-04-2009 Davide Eynard
16. 16
Semantic Wikis
Wikis are not all the same:
Wikipedia + OMP Intranet ?
Generic Internet users who Generic Internet users who Employees with specific
Who write for everyone write for everyone addressee (group, boss)
Internal atomic docs,
Encyclopedia, divided in Hypertextual book (mostly
What articles atomic)
reports, attachments,
annoucements
Contributive, often
Contributive, collective, Mostly collective, explicit
How explicit participation participation
collective, explicit/implicit
participation
Asynchronous (stronger need
When Asynchronous Asynchronous
for updates!)
Centralized in an intranet
Centralized (what about
Where China?)
Centralized context (authorship,
confidentiality)
Social/personal incentives, Social/personal incentives, “Because my boss says so”,
Why small COP often just individuals groups/labs as COP
Semantic extensions:
for data (templates for contents – describe knowledge)
and metadata (about pages and attachments – describe and infer)
Politecnico di Milano, 07-04-2009 Davide Eynard
17. 17
Folksonomies
Main limits: Generic Internet users who
no synonym control Who mostly tag for themselves
“basic level” variations What Classification technology
Contributive, implicit, low
lack of precision How coupling
lack of recall When Asynchronous
lack of hierarchy Where Usually centralized
system-dependent
Why Personal incentives
Folksonomy + Ontology
disambiguate words
add hierarchy
interlink folksonomies
(describe and link)
[B.2, B.3, B.5]
Politecnico di Milano, 07-04-2009 Davide Eynard
18. 18
Annotation Systems: Linking Open (meta)Data
Linking Open Data is:
a project which aims at making data freely available to anyone,
setting RDF links between items from different data sources
a paradigm to publish information on the Web
Metadata = information about information
anything which has a URI can be commented (or annotated)
common limits of annotation systems:
• low user incentives
• cannot reuse metadata over different systems
LOD as a possible solution
• provides a large data base to bootstrap system
• suggests a widespread, machine interpretable standard for
information sharing and reuse
• what about linking other types of data?
[D.1, B.4, D.2]
Politecnico di Milano, 07-04-2009 Davide Eynard
22. 22
Conclusions
we chose a topic
we extended it
(in different directions)
Politecnico di Milano, 07-04-2009 Davide Eynard
23. 23
Conclusions
we chose a topic
we extended it
(in different directions)
this allowed us to study it from other perspectives
Politecnico di Milano, 07-04-2009 Davide Eynard
24. 24
Conclusions
we chose a topic
we extended it
(in different directions)
this allowed us to study it from other perspectives
Results
the behavior and the success of a
participative system depend on many
different factors (i.e. community, incentives,
context)
we developed a general methodology to
evaluate participative systems and
technologies and we applied it to some
specific cases
for each of these cases, we used
semantics to build better tools
Politecnico di Milano, 07-04-2009 Davide Eynard
25. 25
Conclusions
Our contributions
General
• our methodology does not depend on a particular system
Specific
• we applied semantics on different levels (and not just on contents)
to build a better intranet wiki;
• we applied semantic disambiguation techniques to address some of
the most important problem of folksonomies and provided a new,
hierarchical interface to browse tag-based systems;
• we suggested a solution to the bootstrap problem, which employs
already available open data sources to provide a user incentive;
• by converting email and web history to a standard and shared
representation, we automatically linked two of the most accessed
data repositories to a huge amount of related information.
Limits
Our evaluations were mostly focused on datasets and
algorithms
Politecnico di Milano, 07-04-2009 Davide Eynard
26. 26
Conclusions
Future work
Develop a model for the evaluation, the extraction, and the
reconciliation of data coming from different and heterogeneous
data sources
Delve deeper into the intelligent part of collective intelligence
Politecnico di Milano, 07-04-2009 Davide Eynard
27. 27
The end
Thank you! Questions?
Politecnico di Milano, 07-04-2009 Davide Eynard
28. 28
Relevant publications
[D.2] Eynard, D. (2008) Using semantics and participation to customize
personalization. HP Labs Technical Report HPL-2008-197.
[B.5] Bindelli, S., C. Criscione, C.A. Curino, M.L. Drago, D. Eynard, & G.
Orsi (2008). Improving Search and Navigation by Combining Ontologies
and Social Tags. Proc. of the 1st International Workshop on Ambient Data
Integration, Monterey, Mexico.
[B.4] Eynard, D., & M. Colombetti (2008). Exploiting user gratification for
collaborative semantic annotation. Proc. SWUI 2008.
[B.3] Laniado, D., D. Eynard & M. Colombetti (2007). Using WordNet to turn
a folksonomy into a hierarchy of concepts. Proc. 4th Fourth Italian Semantic
Web Workshop, 192–201.
[B.2] Laniado, D., D. Eynard & M. Colombetti (2007). A semantic tool to
support navigation in a folksonomy. Proc. 18th Conference on Hypertext
and Hypermedia (ACM Press, New York), 153–154,.
[D.1] Eynard, D., J. Recker & C. Sayers (2007). An IMAP plugin for
SquirrelRDF. HP Labs Technical Report HPL-2007-161.
[B.1] Eynard, D. (2007). Research on collaborative information sharing
systems. Proc. Knowledge Web PhD Symposium 2007, 81–82.
Politecnico di Milano, 07-04-2009 Davide Eynard
30. 30
Semantic Wikis
Data
• semantic templates using a domain ontology
Metadata
• automatic extraction and ontology supported management of
attachment metadata
• context ontology to describe properties and relations between
documents
Politecnico di Milano, 07-04-2009 Davide Eynard
31. 31
Annotation Systems: Linking Open (meta)Data
Email
• An automatic IMAP to RDF translation tool
• Information can be queried on the fly with
SPARQL, saved in RDF and piped to
external services
Browser history
• Linking history data (visited or bookmarked
URLs) with related, already available
metadata from Freebase
Semantic Annotation Tools
• Exploiting user gratification for collaborative
semantic annotation
Politecnico di Milano, 07-04-2009 Davide Eynard
32. 32
Linking Open (meta)Data
Browser
Linking history data (visited or bookmarked URLs) with related,
already available metadata
Politecnico di Milano, 07-04-2009 Davide Eynard
33. 33
Linking Open (meta)Data
Semantic Annotation Tools
Exploiting user gratification for collaborative semantic annotation
Politecnico di Milano, 07-04-2009 Davide Eynard
35. 35
Taxonomy of participation - a total mess!
alone with others
do/create
suggest
share
contributive
participative/collective
social
Politecnico di Milano, 07-04-2009 Davide Eynard
36. 36
Levels of participation - The county fair example
• many people participate (take part) to the event
• some contributed with suggestions, actions or resources
• some collaborated to organize it
• some are just there to have fun, or for their business
• ... but the fair success depends on all of them!
• see Engeström's Activity Theory
Politecnico di Milano, 07-04-2009 Davide Eynard
38. 38
Folksonomies: datasets and stats
x axis: tags ordered by usage
y axis: percentage of tags belonging to WordNet
We have collected a large
dataset from del.icio.us that
allowed us to study tags
statistically.
Politecnico di Milano, 07-04-2009 Davide Eynard