Ambient Intelligence and Social Awareness

Download Report

Transcript Ambient Intelligence and Social Awareness

Towards Decentralized
Communities and
Social Awareness
Pierre Maret
Université de Lyon (St Etienne)
Laboratoire Hubert Curien
CNRS UMR 5516
1
Who I am?
Pierre Maret
 PhD in CS (1995)
 Ass. Prof. at INSA Lyon (1998-2007)
 Prof. at Univ of St Etienne (Univ. of
Lyon) since 2008
 Research background : DB, IS,
electronic documents, knowledge
management, knowledge modeling
2
Talk on:
 Towards Decentralized Communities
and social Awareness
3
A Community ?
 What is it?




A
A
A
A
set of participants?
topic?
protocol for the exchange of messages?
data base for storing some information?
 Actually, what is/are the objectives?
4
Improve information exchanges
 Increase efficiency
 Create new opportunities for relevant
exchanges
 Enable exchange of new types of
information
 Deliver the right information, at the
right moment, and to the right person
5
Domains addressed
 Knowledge modeling
 Information diffusion, sharing, retrieval
 Recommendation systems
6
Social Networks Sites
 Great success
 4 types:




Content Sharing (i.e. U-Tube)
Social Notification (i.e. Facebook)
Expertise Promotion (i.e. Wikipedia)
Virtual life, games (i.e. Second life)
 Great tools for building communities
7
Social Networks Sites
 Regarding Content sharing and Social
notification:
People trust people they know
Social network ↔ Decision making
Decision making =
 to follow recommendations
 to imitate behavior
 to support in real-life activities
8
Social Networks Sites
 Social networks can be useful
 but SNS have some drawbacks
9
Some drawbacks of SNS




Multiple registration
Close world (no interoperability)
Privacy issues
No control on data deletion
 Towards a unique governmental
secure SNS ? No
 Then what?
10
Need for an open approach
 An open approach for communityrelated information exchanges
 include interoperability
 avoid personal data dispersion
 Proposal: A community abstraction
Decentralized
+ bottom-up approach
11
Towards a decentralized approach
 1st step : Actors
 2nd step : Communities
 3rd step : Context
12
Towards a decentralized approach
 1st step : Actors
 Actors : an abstraction to model any
participant
 Person
 Personnel assistant (artifact)
 Autonomous system (artifact)
 An actor has
 Knowledge
 Behavior (decision abilities, actions)
13
Actors as SW agents
 2 types of agents:
 Context agent
 Dedicated to sensors
 From raw data to information
 Personal agent
 Personal assistant. Pro-active (internal goal)
 Contains some user's knowledge
 Knowledge is "delivered to" and
"gathered from" the environment
 Mobility scenario or in-office scenario
14
Personnel agent
 Role of a user assistant
 Piece of software
 Autonomous software with communication
abilities
 Knowledge = abstraction of the owner's
knowledge
 Decision abilities = actions (managed by the
owner), related to the present knowledge
15
Actor abstraction
{ ki } knowledge
{ bi } behavior
Actor
Actor
{ ki } knowledge
{ bi } behavior
Actor
{ ki } knowledge
Tulip is_a Flower
Red is_a Color
Tulip has_property Red
T1 instance_of Tulip
{ bi } behavior
Send message
Receive message
Extract Instances
Set Value
 Expressed using web semantic techniques :
OWL
16
Making behavior exchangeable
 Knowledge (RDF/OWL ontologies) can be
exchanged
 Behavior is generally hardcoded : not
exchangeable
 A model for expressing agent's behavior in
SWRL (expression of rules on OWL)
 Work of Julien Subercaze (PhD candidate)
17
Making behavior exchangeable
 Behavior as a finite state machine
 If (transition from State A to State B)
then (execute list of actions)
18
Describing information
 Using Tags to describe agents
information/knowledge
 Tag = Annotations, Meta-data
 Concerns any
information/knowledge/document
 picture
 signal
 email, etc.
19
Tagging activity on personal agents
 Tagging activity
 Automated
 Semi-automated
 Manual
 Useful regarding information retrieval
 Several dimensions/processes for tags
 Location, environmental information, body
information, thoughts, …
20
Tagging activity on personal agents
 Work of PhD candidate Johann Stan
 Main idea : the meaning of tag
changes dynamically according to the
user and circumstances.
 Circumstance :
 communities the user belongs to
 context
21
2nd step : Communities
 1st Step : Actors
 Community : A set of actors with compatible
communication abilities and shared values
(common domain of interest)
 VKC = Virtual Knowledge Communities
An abstraction for the exchange of information inbetween actors
22
Features for communities
 Community-related knowledge of the agents
 List of (some) communities
 List of (some) agents
 Community-related domain knowledge (about the
community topic)
 Community-related primitives
 Protocol: create, inform, request…
 Knowledge selection (extract from its knowledge)
 Knowledge evaluation and insertion (received
through exchanges)
23
Features for communities Communities
Knowledge
Mappings
24
Agent communities
 Community protocol
 Create community (with a topic)
 Join, Leave
 Inform, request
 Specific role (any agents)
 Yellow page
 Knowledge = existing communities and
topics
25
Example
A1
{ ki }
//joint communities
C1 (on Car)
C2 (on Flower)(Owner)
{ ki }
Tokyo is_a City
//joint communities
C1 (on Car)
A3
A3
A2
A2
A3
{ ki }
Tulip is_a Flower
C1 is a Community
C2 is a Community
//joint communities
C2 (on Flower)
A2
has previously joined A1's community on Flowers.
wants to send some info to this community
needs more info about Japan.
is about to create a community on Japan
26
Communities and social network
 Memory of interactions builds my social
network




With who?
The topic?
The context?
The environment?
 Carried out with tags
 Used to propose interaction facilities
(prediction)
27
Communities and social network
 Example of annotations of interactions
(manual)
 Automatic annotations: context, content analysis
 More about the context…
28
Step 3 : Context
 Context data: gathered from the environment




Location
Internal state
Environment
Activity (…)
 Situation = f(context data)
 SAUPO model:
situation ↔ communication preferences
29
SAUPO model
Situation ↔ Communication preferences
30
Agent's context
 User's current activity as context data
 Identifying the user's current activity to
promote exchanges
 Event + Content analysis and filtering
 Target : more accurate solicitations
 Contextual Notification Framework
31
Agent's context
 Contextual Notification Framework (Work of
Adrien Joly, PhD Candidate) Filtered
ambient awareness
 Main idea :
 maintain cooperation in-between people
 while reducing overload
 Context model
 Context sniffer (with user acceptance)
 Matchmaking process (context + social
network) and notification
32
Contextual Notification Framework
33
Conclusion
 Improving knowledge exchanges
 Used techniques
 Semantics modeling: ontologies, owl
 Context awareness
 Social networks
 Leveraged into several scenarios or
projects
 Leading idea : bottom-up approach
34
Thank you for your attention
35