Upload
elwin-stafford
View
223
Download
0
Tags:
Embed Size (px)
Citation preview
2/Presentation Outline
• Introduction
• Motivation
• Related Work
• Limitations
• Proposed Architecture
• Tools and Technologies
• Development Timeline
• Current Status
3/Introduction
• According to Dey, Abowd and Salber (2001)• Context is any information that can be used to characterize the situation of
entities (i.e. whether a person, place or subject) that are considered relevant to the interaction between a user and an application, including the user and the application themselves
• Means Context is very dynamic and transient
• Low-level Context is the raw data coming from sensors / source
• High-level Context is the abstract information extracted from logically relevant low-level context
4/Motivation
• End-users are interested in high-level context, not raw data• For example“It’s raining” (high-level) is preferred rather than “humidity: 77%“ (low-level)
• To discover Higher-level Context Information• Context fusion is required
• Find the semantic relationship
• By integrating all the information relevant to one’s life• It can give us a real sense of Life logging
5/Motivation
• Context Information• Sensor data from sensors• Activity data from smartphones• Nutrition data• Social media data
• Recommending Services• Lifestyle & Lifecare• Exercise and nutrition • Social interaction• Behavior analysis &
prediction
Behavior Modeling
Lifestyle Analysis,
Prediction, andRecommendati
ons
Context Informatio
n
Context Informatio
n
Context Informatio
n
6/
User Profi
le
Social
Media
Life log
Environmen
t
Proposed Idea
IntegrationSocial MediaInformation
Diet Information
Activity Information
ProfileInformation
Environment Information
Beh
avio
r
Mod
elin
g
Activity
Context Information Independent logs
Life log Information Semantically related
7/Related Work
Ontology/ Subdomain
CoBrA-Ont[2003]
CoDAMoS[2004]
SOUPA[2005]
Delivery Context [2009]
Device X X X
Environment X X X
Location X X X X
Network X
Role X X
Time X X X
User X X X
Implementation Available
X X X
8/Related Work
Ontology/ Subdomain
CONON[2003]
Situation Ont[2004]
mIO[2005]
PiVOn[2009]
PalSPOT[2011]
Device X X X X X
Environment X X X X X
Location X X X X X
Network X X
Role X X
Time X X X X X
User X X X X X
Implementation Available
X X
9/Limitations
• Domain Specific Context Models
• Information exist in different types of Logs but are stand alone
• No integrated system
• Very few systems consider social interaction
• None of the existing systems use any behavioral model
• High level contexts are extracted using context interpretation and aggregation techniques regardless of
• Context verification, validation and fusion
10/Proposed Architecture
Long/Short-term Behavior AnalysisContext Awareness
Mapper and Transformer
Context Interpreter
Context Analyzer
Rulebase
Parser
Decision Making
DTD2OWL OWL Ontology XML2OWL
QueryGenerator
ActivityRetrieval
Match Making
Rule-basedFiltering
Decision PropagationSituation Analyzer
Prediction and ReasonerRecom
mendation M
anager
Low Level Context-awareness
HDFS Data Interface
Intermediate Data
Life log Modeling
Context Receiver
Context Representation
Context Verification
Context Fusion Context Logging
Behavior Modeling
ParserLife Log
Repository
Feature Selection
Pattern Classification
Pattern Identification
Life log Extractor
11/ScenarioContext Awareness | Long/Short-term Behavior Analysis
Prediction and Reasoner Recommendation Manager
Low Level context-awareness
HDFS Data Interface
Intermediate Data
Long/short termBehavior Analysis
Mapper and Transformer
Context Interpreter
Context AnalyzerParser
Decision Making
DTD2OWL OWL Ontology
QueryGenerator
ActivityRetrieval
Match Making
Rule-basedFiltering
Decision PropagationSituation Analyzer
7
4
2
11
6
7
3
Rulebase
5
7
Activities detected: walkingLocation detected: KitchenTime noticed: 12:00:00
Ali Exercising Sentiment: tired
Rule 1 (activity=eating AND time<8:00:00) Taking BreakfastRule 2 (activity=eating AND time=12:00:00) Taking LunchRule 3 (activity=eating AND time=18:00:00) Taking Dinner
<activity>eating</activity><activityLocation>Kitchen </activityLocation><activityTime>12:00:00 </activityTime><performedBy>Ali </performedBy>
"SELECT ?activityName ?hasConsequentAction ?type ?performedBy ?time WHERE { <" + strNS + strActivity + "> <" + strNS + "hasName> ?activityName ." +"<" + strNS + strActivity + "> };
In Kitchen Eating 12:00:00
Taking Lunch in Kitchen
Rule 2 Matched(activity=eating AND time=12:00:00) Taking Lunch
XML2OWL
12/ScenarioContext Awareness | Long/Short-term Behavior Analysis
Context Receiver
Context Representation
Context Fusion
Context Verification and LoggingLife Log Repository
Behavior Analysis
Life Log Extractor
Parser
Context ConverterContext Representation Context Mapper
Horizontal Context Fusion
Vertical Context Fusion
DataExtractor
Data Logger
Query Formulation
Data Fetcher Data Processing
Pattern Classification Pattern Identification Feature Selection
Consistency Verification
SemanticStructural
Log ContextExistence Verification
Low Level context-awareness
High Level Context-awareness
1 1 HDFS Data Interface
Intermediate Data
1
Prediction and Reasoner Recommendation Manager
3
10
9
8
7
6
5
4
10
2
Verified Data
<activity>eating</activity><activityLocation>Kitchen</activityLocation><activityTime>12:30:00</activityTime><performedBy>Ali</performedBy>
Lunch, Exercising
Complete Information of User like profile, activities performed when and where, tweets, etc
Behavior AnalysisLunch No proper timingExercising Regular
High Level Context: Exercising Structured Data
Ali,Lunch,Hotel,2014-05-02, 12:00:00
walking, sitting, eating, tweet: tired
Activity detected: walking, eatingLocation detected: KitchenTime: 12:30:00
13/Tools and Technologies
• Knowledge Representation–RDF / RDFS / OWL Ontologies–Protégé as IDE
• Programming Language & API– Java– Jena, Twitter
• Querying Language–SPARQL
• Reasoner–Racer Pro / Pellet / FaCt++
14/Development Time LineFour Year Work Break Down
Literature Study:Context Modeling
Analytical Study:Context Conversion
Analytical Study:Context Interpreta-
tion
Literature Study:Life log System
Context logging in Life log
Context Integration& Verification
Literature Study:Behavior Representa-
tion
Study:Behavior Modeling
Algorithmic Study:Behavior Analysis
System Develop-ment
Unit Testing
Integration Testing
Context Model
Context MappingTechnique
Context Analyzer
Life log Model
Life log Parser
Fusion + VerificationTechniques
Features SelectionBehavioral Model Cre-
ation
Model SelectionPattern Identification
Algorithm: Long-term /
Short-term Analysis
High-levelContext Awareness
Test Report
Test Report
First Year Second Year Third Year Fourth Year
Context awareness Life log Management Behavioral Analysis Documentation
Prototype Develop-ment
Prototype Develop-ment
Prototype Develop-ment Technical Guide
15/Current Status
• Context Awareness– Literature Survey regarding
– Context Modeling Techniques– Context Conversion & Mapping Algorithms
• Long/Short-term Behavior Analysis – Literature Survey regarding Lifelog Design and Development
• LifeLog Model: OWL+Dynamic/Static• Storage: Ont-RDB / JenaTDB / RDF• Nature & type of Logs
21/Proposed Architecture
Prediction and Reasoner Recommendation Manager
Low Level context-awareness
HDFS Data Interface
Intermediate Data
Long/Short-termBehavior Analysis
Context Awareness
22/Proposed Architecture
Prediction and Reasoner Recommendation Manager
Low Level context-awareness
HDFS Data Interface
Intermediate Data
Long/Short-termBehavior Analysis
Mapper and Transformer
Context Interpreter
Context Analyzer
Rulebase
Parser
Decision Making
DTD2OWL OWL Ontology XML2OWL
Query Generator
Activity Retrieval
Match Making
Rule-based Filtering
Decision PropagationSituation Analyzer
Context Awareness | Long/Short-term Behavior Analysis
23/Proposed Architecture
Low Level context-awareness
HDFS Data Interface
Intermediate Data
Context-Awareness Context Receiver
Context Representation
Context Fusion(Vertical / Horizontal)
Context Verification and Logging
Life Log Repository
Behavior Analysis
Life Log Extractor
Parser
Prediction and Reasoner Recommendation Manager
Context Awareness | Long/Short-term Behavior Analysis
24/Proposed Architecture
Long/Short-term Behavior AnalysisContext Awareness
Mapper and Transformer
Context Interpreter
Context Analyzer
Rulebase
Parser
Decision Making
DTD2OWL OWL Ontology XML2OWL
QueryGenerator
ActivityRetrieval
Match Making
Rule-basedFiltering
Decision PropagationSituation Analyzer
Prediction and ReasonerRecom
mendation M
anager
Low Level Context-awareness
HDFS Data Interface
Intermediate Data
Lifelog Modeling
Context Receiver
Context Representation
Context Verification
Context Fusion Context Logging
Behavior Modeling
ParserLife Log
Repository
Life Log Extractor
Behavior Checker
Model CreaterBehavior Descriptor
Model Validator
Behavior Analyzer
25/Motivation
• To integrate the different context information emerging from diverse sources to identify user’s behavior in order to analyze the user’s lifestyle and provide recommendations to promote active lifestyle
Social MediaInformation
Diet Information
Activity Information
ProfileInformation
Environment Information
Lon
g / S
hort-te
rm
Beh
avio
r Mod
elin
g
Life log