SAMOA: Sociall y-Aware School of Information Sciences - University of Pittsburgh MObile Architecture Pittsburgh, PA, Feb. 27th, 2009 Dario Bottazzi [email protected]DEIS—University of Bologna Viale Risorgimento 2, 40136 Bologna, Italy Outline Outline Social Computing New Opportunities for Social Computing in New Opportunities for Social Computing in Pervasive Environments The SAMOA Framework – The SAMOA Model – The SAMOA Architecture University of Pittsburgh — Feb. 27th, 2009 Experimental Results
15
Embed
SAMOA: Socially-Aware MObile Architecturepitt.edu/~dtipper/2011/Slides4_09.pdf · The SAMOA Framework Socially-Aware MObile Architecture – Supports the creation of any-where, any-time
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
SAMOA: Socially-Aware
School of Information Sciences - University of Pittsburgh
DEIS—University of BolognaViale Risorgimento 2, 40136
Bologna, Italy
OutlineOutline
Social Computing
New Opportunities for Social Computing inNew Opportunities for Social Computing in Pervasive Environments
The SAMOA Framework – The SAMOA Model
– The SAMOA Architecture
University of Pittsburgh — Feb. 27th, 2009
Experimental Results
Social NetworksSocial Networks
o man is an island entire of itself; every man isNo man is an island, entire of itself; every man is a piece of the continent, a part of the main.N
John Donne, Devotions Upon Emergent Occasions—Meditation XVII
University of Pittsburgh — Feb. 27th, 2009
The Roots of Social ComputingThe Roots of Social Computing
Since from early 60s several researches in Computer Supported Cooperative WorkComputer Supported Cooperative Work (CSCW) investigated suitable means to leverage human connections and improvinghuman interaction– Interaction driven by business needs
University of Pittsburgh — Feb. 27th, 2009
– Structured and Formalized user collaboration
Social ComputingSocial Computing
Recent times several Internet-based social computing supports have emerged in the literature– Automatic Identification of experts in a field
– Study of Social Dynamics in Communities (e.g. for team-building)
– Improvement of Web Ranking Algorithms
– Sharing and Distribution of Pictures, Videos and text messages
University of Pittsburgh — Feb. 27th, 2009
messages
– …
Social ComputingSocial Computing
Internet is assumed to be the main source of empirical data from which to extract meaningful social relationb t i di id lbetween individuals– Links on Web Pages– Scientific Publication Data-Bases– Repositories of e-mail messages (e.g. in enterprise settings)– Blog or Usenet posts– …
University of Pittsburgh — Feb. 27th, 2009
Data mining techniques are often employed to extract the social network. Only recently some ontology-basedapproach for modeling and representing meaningful social relation has appeared (e.g. FOAF profiles)
Social ComputingSocial Computing
OpenSocial API: de-facto industrial standard for the development of social-computing Web sites– Define a set of API that determine the set of
mechanisms needed to build and update socialnetworks
– Allow different JavaScript-based applications (widget) to access the social network of a user to promoteand support informal and swarming collaboration
University of Pittsburgh — Feb. 27th, 2009
pp g– Widget can be developed by third-parties and can be
used within each open-social compliant web site– Apache Shinding
New New Opportunities for Social Opportunities for Social Computing Computing
The Pervasive Scenario– Proliferation of wireless-enabled devices
– Recent advances in Wireless Networks and the emergence of MANET technology
Pervasive Computing poses challenging issues to the Social Computing research circles
– What kind of mobile social application can we build?
University of Pittsburgh — Feb. 27th, 2009
– What are the new requirements of mobile social applications?
– How can we support the development of mobile social applications?
Impromptu Collaboration (1)Impromptu Collaboration (1) A set of co-located autonomous entities that
can impromptu communicate, collaborateith h thwith each other
– Collaboration occurs between individuals who sharecommon interests, activities, tasks and goals
Collaborating parties interact to mutuallyprovide each other services and resources o
University of Pittsburgh — Feb. 27th, 2009
provide each other services and resources otheir reciprocal interest
Highly Dynamic Nature of MANETs– Frequent Host Connections and Disconnections
– Frequent Network Partitions and Merges
– Wireless medium-related issues
– Hidden/exposed terminal problem, grey zones, …
Heterogeneous Characteristics of Collaborating P ti
University of Pittsburgh — Feb. 27th, 2009
Parties– Impossibility to make assumption on user devicenature and capabilities
– Resource scarcity
Our ApproachOur Approach
Available Social Computing Application for Pervasive Environments
B ilt t f N t k L– Built on top of Network Layer
– Provide Support Only for Specific Applications
– Hard to re-use in different scenarios
– Love-Gety, Proxy-Lady, Social-Net, …
Need for a middleware solution to
University of Pittsburgh — Feb. 27th, 2009
– Address social network management details (e.g., matchmaking support for social network formation, representation of user characteristics, …)
– Address technical challenges (e.g., heterogeneity in terminals, disconnections, user location/tracking…)
The SAMOA FrameworkThe SAMOA Framework Socially-Aware MObile Architecture
– Supports the creation of any-where, any-time semantic socialnetwork that groups together mobile users in reciprocalnetwork that groups together mobile users in reciprocal physical proximity and show commonalities (e.g. similar interests, activities)
– Social networks in SAMOA are centered around the user (ego-centric social network)
–Social networks reflect all past and present socially-relevantencounters of the user
University of Pittsburgh — Feb. 27th, 2009
encounters of the user
Context-Awareness in SAMOA– Place-Awareness: visibility of the place where the user is located
– Profile-Awareness: visibility of place activities, and user attribute, interests, activities, …
The SAMOA ModelThe SAMOA Model
1 hop2 hops
University of Pittsburgh — Feb. 27th, 2009
2 Hops Locality1 hop
2 hops
The SAMOA ModelThe SAMOA Model
Place 1Place 2
Manager 1
Manager 2Client 1 Client 2
Client 3
Client 4
University of Pittsburgh — Feb. 27th, 2009
2 Hops Locality1 hop
2 hops
Metadata ModelMetadata Model
<profile:Place <profile:User rdf:ID=“Alice”>
<profile:Userrdf:ID=“BookVendor”>Id
rdf:ID=“Bookshop”>...<profile:id>
...
<profile:activity>...
Ac
tivi
tyId
...<profile:_id>...
<profile:activity>...
<profile:act pref>
Ac
tId
ef
...<profile:id>...
<profile:preference>...</profile:User>
Pre
fere
nce
Discovery Profile
University of Pittsburgh — Feb. 27th, 2009
</profile:Place><profile:act_pref>...
</profile:User>
Pre
Place ProfileUser Profile
Discovery Profile
Social Network Extraction Social Network Extraction ModelModel
Semantic social matchmaking algorithms can recognize different semantic relationships on the basis of the subclass relationships defined inbasis of the subclass relationships defined in activities/preferences ontologies. User activities and preferences can be
– an instance of the activity/preference class in manager’s place or discovery profile (exact case)
an instance of a more generic activity/preference class in
University of Pittsburgh — Feb. 27th, 2009
– an instance of a more generic activity/preference class in manager’s place or discovery profile (subsumes case)
– an instance of a more specialized activity/preference class in manager’s place or discovery profile (plug-in case)
Social Networks in SAMOASocial Networks in SAMOA
Place-dependent Social Network
– reflects the set of members of manager’s social network in the same placesame place
Global Social Network– persistently records the whole set of place-dependent social networks dynamically created over time as the manager moves across places
University of Pittsburgh — Feb. 27th, 2009
SAMOA ArchitectureSAMOA Architecture
Socially-Aware Applications
SAMOA Social Network Management LayerSAMOA Social Network Management Layer
Profile Manager (PM)Semantic Matching
Engine (SME)
Place-dependentSN Manager (PSNM)
GlobalSN Manager (GSNM)
SAMOA Basic Service Layer
University of Pittsburgh — Feb. 27th, 2009
JVM-OS-HW-Network
Message Transport Manager
Location/Proximity Manager
SAMOA Basic Service Layer
SAMOA Service InteractionSAMOA Service Interaction
PSNM SME PM L/PM MTMGSNM PM SMEL/PMMTM
Manager Client
PSNM GSNM
Send placeadvertisement
Broadcast placeadvertisement
Notify placeavailability
Forward placeadvertisement
Send entityadvertisement
Broadcast placeadvertisement
Forward entityadvertisement
PP RequestForward PPrequest
Notify PP request
Return PP Forward PP Forward PP
Profile filteringrequest
Returnfiltered UPprofile
Filter UPaccording toreceived PP
Forward filtered UPForwardfiltered UP
Forwardfiltered UP
Verify semanticcompatibility between
li tÕ UP d
Profilematchingrequest
Profile
University of Pittsburgh — Feb. 27th, 2009
clientÕs UP andmanagerÕs DP
Profilematchingresponse
Include Client in bothPlace-dependent andGlobal Social Networks
Social Matchmaking Algorithm Social Matchmaking Algorithm EvaluationEvaluation
Settings
– 64 entities. Up to 2 activities in user/discovery profiles
Recall– the extent to which all socially related individuals are included in the network (by avoiding false negatives)
–Recall is optimal because our algorithm is complete
Precision
University of Pittsburgh — Feb. 27th, 2009
– the extent to which only socially related individuals are included in the network (by avoiding false positives)
– Good level of precision, because the algorithm can look for manager specified preference values only in the semantically correct activity type, thus reducing false positives
ConclusionsConclusions Supporting Social Computing in Pervasive
Environment is a challenging task
Middl l l t ff i t ti Middleware-level support can offer interesting opportunities to leverage the development of new social computing services and applications
We are currently validating our support in different application scenario
University of Pittsburgh — Feb. 27th, 2009
pp
We are starting to investigate security concerns stemming from social computing applications for pervasive environments
SAMOA DistributionSAMOA Distribution
Freely available upon request
S OThe distribution includes the SAMOA middleware along with an emulation environment that simplifies development and experimentation of social applications in Pervasive Environments
We are more than happy to share our solution within the
University of Pittsburgh — Feb. 27th, 2009
research community
AcknowledgementsAcknowledgements
University of Pittsburgh — Feb. 27th, 2009
FIRB TOCAI.IT Project “Knowledge-Oriented Technologies for Enterprise Aggregation in