Mobile social search
-
Upload
ramesh-jain -
Category
Health & Medicine
-
view
107 -
download
1
description
Transcript of Mobile social search
9/25/12 1
An Elephant and Six Blind Men
An Elephant is NOT
• Wall • Rope • Snake • Spear • Tree • Fan
An Elephant is … an Elephant
Mobile Social Search (MSS)
is NOT Mobile is NOT Social is NOT Search
Is Mobile Social Search
Changing Times in Search
Mobile Social Search: Search ‘inside’ problem solving in real ;me.
Problem Solving
Search
MSS: A Problem of Riches
When we were (data) poor – we searched.
Now that we are (data) rich – we need mobile social search.
Search
Connecting
Then
Now
Then
Now
• What is an Italian Restaurant? – List Italian Restaurants in Brussels.
• Show me Italian Restaurants? – Use Yelp (or any other social approach) and distance from me to rank them.
• Where should I go to eat Italian food NOW? – Consider Tme taken, current ambiance, and quality of food into consideraTon.
• What are symptoms of Swine Flu? • Is there a major Swine Flu outbreak in my area? • You are likely to be very sick with extreme case of Swine Flu soon, your doctor is ready with the set up.
Social Networks
Connecting
People
Connecting People
And Resources
Social Life Networks
Aggregation and
Composition
Situation Detection
Alerts
Queries
Information
9/25/12 14
Concept recogni;on from data
15
Environments
Real world Objects
SituaTons
AcTviTes
Heterogeneous Media
Single Media
SPACE TIME
Scenes LocaTon aware
Visual Objects
Trajectories
Visual Events
LocaTon unaware
StaTc Dynamic
Heterogeneous Media
Single Media 360 K 11.4K
3.4 K LocaTon aware
LocaTon unaware
StaTc Dynamic
9/25/12 Proprietary and ConfidenTal, Not For DistribuTon 16
• SituaTon: An acTonable abstracTon of observed spaTo-‐temporal characterisTcs.
• Allow users to define their own spaTo-‐temporal features and create the situaTon detecTon filters.
9/25/12 17
Swine flu: Situa;on Segmenta;on
into ‘high’ and ‘low ’ac;vity zones.
9/25/12 18 Proprietary and ConfidenTal, Not For DistribuTon
(a) Pollen levels (Source: Visual) (b) Census data (Source: text file) (c) Reports on ‘Hurricanes’ (source: Twitter stream)
d) Cloud cover (Source: Satellite imagery) (e) Predicted hurricane path (source: KML) (f) Open shelters coverage(Source: KML)
Representa;on for different data sources into a common spa;o-‐temporal format.
Level 1: Unified representaTon (STT Data)
Level 3: Symbolic rep. (SituaTons)
ProperTes
ProperTes
ProperTes
Level 0: Raw data streams e.g. tweets, cameras, traffic, weather, …
Level 2: AggregaTon (Emage)
…
STT Stream
Emage
SituaTon
9/25/12 21
Billions of data sources. Selec;ng and combining appropriate sources to detect situa;ons. Interac;ons with different types of Users
Decision Makers Individuals
Front End GUI
NewDataSource
NewQuery
E-‐mageStream
E-‐mage Stream
E-‐mage Stream
Data Cloud
Back End Controller
Stream Query Processor
Data IngestorRegisteredData
Sources
RegisteredQueries
Raw Spatial Data Stream
API Calls
Raw DataStorage
Personalized Alert Unit
AlertRequest
User Info
9/25/12 22
9/25/12 23
S. No Operator Input Output 1 Selection σ Temporal
E-‐mage Set Temporal E-‐mage Set
2 Arithmetic & Logical⊕
K*Temporal E-‐mage Set
Temporal E-‐mage Set
3 Aggregation α Temporal E-‐mage set Temporal E-‐mage Set 4 Grouping γ Temporal E-‐mage Set Temporal E-‐mage Set 5 Characterization :
• Spatial φ • Temporal τ
• Temporal E-‐mage Set • Temporal Pixel Set
• Temporal Pixel Set • Temporal Pixel Set
6 Pattern Matching ψ • Spatial φ • Temporal τ
• Temporal E-‐mage Set • Temporal Pixel Set
• Temporal Pixel Set • Temporal Pixel Set
24 9/25/12 24
Personal SituaTon
9/25/12 Proprietary and ConfidenTal, Not For DistribuTon 25
Situa;onal controller
• Goal • Macro SituaTon • Rules
Micro event e.g. “Arrgggh, I
have a sore throat” (Loc=New York, Date=12/09/10)
Macro situa;on
Control Ac;on “Please visit nearest CDC
center at 4th St immediately”
Date=12/09/10
Alert Level=High
Level 1 personal threat + Level 3 Macro threat -‐> Immediate ac;on 9/25/12 26
e.g. High Flu risk
+
1) Macro situaTon
Social sensors
Device sensors
Macro sensors
Personal life streams
Profile/ Preferences
+
2) Personalized situaTon
+
Planetary scale sensing
Personal context
Available resources
3) Recommend AcTons
Resource data
9/25/12 28
Flood level - Shelter
Flood Level Shelter
Classify (Flood level - Shelter)
Taking personalized acTons
9/25/12 29
• Was for Archived data and mostly for researchers (using Desktop).
• Currently: Mobile clients, Local data, and Social Graph (SoMoLo).
• Immediate Future: SituaTons from Real Time data and RecommendaTons.
• Future: PredicTve control of emerging situaTons.
Thanks for your Tme and alenTon.
For quesTons: [email protected]
9/25/12 32
Applica;on scenarios
¨ Business decision making: Demand-‐supply analysis, opening a new store, offer,…
¨ Medical : Epidemic monitoring, Asthma, polluTon effect miTgaTon
¨ Disaster relief: (hurricane, flood, fire) direcTng people to appropriate resources.
¨ Traffic: SuggesTng best routes