Transcript Document
A Pragmatic Approach to Realizing
Context-Aware Personal Services
Sangkeun Lee, Dongjoo Lee, Seungseok Kang, and Sang-goo Lee
School of Computer Science and Engineering
Seoul National University, Seoul
{liza183, therocks, pyxis81, sglee}@europa.snu.ac.kr
ECBS Workshop 2008
Sydney
1
Context-Aware Personal Services
Network and computing technologies
have improved
The number of mobile devices has
increased
The number of content and service
have increased
Context Aware-Personal Service
• Actively and autonomously adapts and
provides the most appropriate service or
content to users, accurately interpreting the
context without much user supervision.
Service
Voice
recorder
Cell phone
Content
PC
Home
Network
GPS
Navigator
PDA
User
Service
Content
User’s Context
Service
Digital
camera
How can we provide the Service
most appropriate service
or content to users?
Content
Content
Realizing context-aware personal services has become one of the most
important issues in pervasive computing
Copyright 2008 by CEBT
Semantic Tech & Context - 2
Context-Aware Personal Services
“In a world of infinite choice, context – not content – is king.”
(Listens.com)
Context
Primary types: identity, location, time, activity
Secondary types: all other sources of information, either explicit or
implicit, that help us better determine the user’s need
Context-aware
adaptive, reactive, responsive, situated, context-sensitive, environmentdirected
Issues
Modeling, Processing, Security & Privacy, Robustness, Scalability &
Performance
Our Objective
To build a practical semantic technology adoption technology for
context-aware personal services
To present a context-aware personal service framework based on our
strategy
Copyright 2008 by CEBT
Semantic Tech & Context - 3
Survey on Context-Aware Computing
Copyright 2008 by CEBT
Semantic Tech & Context - 4
Survey on Context-Aware Computing
Achieving generality and feasibility
at the same time is very difficult
Copyright 2008 by CEBT
Semantic Tech & Context - 5
Practical Semantic Technology Adoption
for Context-Aware Personal Services
Some existing approaches represent as much context information as possible in
ontology
Some previous approaches have attempted to achieve the ideal goal of machines
“understanding” user contexts as well as humans do, and have considered the
limitations of reasoning systems as problems to be solved in the future.
Some existing approaches have relied on complex technologies such as description
logics, artificial intelligence, and natural language processing
Current existing ontology reasoning systems cannot process the high-level inferences
Conventional Approach
Practical Approach
Focus on what we don’t have
upper ontology,
powerful inference engines
Focus on what we have
relational database technology,
transactional data, inference
engines that have limitations
Focus on Inference
complex concept hierarchies
FOL vs. F-Logic vs. DL
Focus on data
simple usable concepts
low degree of inference
Satisfy scientific constraint
completeness, decidability
Satisfy business value
performance, feasibility
Copyright 2008 by CEBT
Acceptable
Performance
& Scalability
Semantic Tech & Context - 6
A Context-Aware Personal Service Framework – Conceptual
Architecture
Our framework is designed based on the strategy presented in previous section
To enable a flexible and easily extensible context-aware personal service, it is necessary to
consider a wide range of issues
–
Hardware & Network Infrastructure
Here, we focus on software technology
Copyright 2008 by CEBT
Semantic Tech & Context - 7
이상근
Context-Aware Personal Service Process Flow
Context Manager controls the
abstraction levels of context
information based on predefined
context rules
Request Processor chooses an
Dynamic
Info.
(Activity)
Static Info.
(Profile)
Users
Service Providing Define
effective algorithm to provide the
appropriate content or service,
and performs actual services.
Contents Recommendation
Automatic Device
Configuration
Seamless Service
Home
Network
PDA
PC
Digital
camera
GPS
Navigator
Device-dependent Context Data (Low-Level)
Context
Manager
Context
Rule
Integrated Context Data (High-Level)
Service Trigger
Service
Rule
Request
Request
Processor
Define
Applications
Voice
recorder
Use
Service Trigger identifies a
match between current context
match and a predefined service
rule, then triggers a request to
the Request Processor
Cell
phone
Select
Ranking
Algorithm
Support
Rule Mining
Service Providers
Data Mining
Copyright 2008 by CEBT
User
DB
Music
DB
Movie
DB
Log
Large amount of
video & audio resources
in Legacy Database
Semantic Tech & Context - 8
A Context-Aware Personal Service Framework –
Layer Separation
Service Request Layer
focuses on context information
The system controls the abstraction levels of much low-level
information and transforms it into manageable numbers of high-level
context information.
Our reasoning engine performs inferences with a reasonable amount
of context information, thereby increasing performance in this layer.
Service Processing Layer
manages a number of large-sized content databases such as music
and movie databases
We assume that the semantics of the content data are relatively less
important than the context information and thus use relational
database technology.
However, simple inferences can still be processed using this method
by maintaining well-defined and clean data.
Copyright 2008 by CEBT
Semantic Tech & Context - 9
Implementation
We implemented the prototype of our framework, named
CAP(Context-Aware Personalization) in Java
XML-RPC protocol
Context and Service rule in XML format
Simple reasoning engine
Simple context data model
Several demo scenarios
Virtual devices
Copyright 2008 by CEBT
Semantic Tech & Context - 10
Use Case
We were able to construct useful demo scenarios using CAP
A is driving his car to the airport. As soon as the fuel-alert indicator lights,
the system matches the current context to predefined Gas Station
Recommendation Service. Then it triggers Gas Station Recommendation
Service, and sends the context information, including location, route, time,
and user identification, to the proper service provider. Gas Station
Recommendation Service Provider searches the best 10 gas stations,
using the user profile data that they have and context that they received
from CAP. Then the service provider sends the list of gas stations to the
system, which displays the list on the interface in A’s car
Acceptable Scalability & Performance because
We do not allow complex reasoning or ontology
We use simple context-data & rule model
We process large number of instances using RDBMS
Copyright 2008 by CEBT
Semantic Tech & Context - 11
Conclusions & Future Work
Conclusions
We presented a strategy with respect to conventional technologies, such as
relational databases.
We emphasized well-defined and clean data, rather than inference itself and
mathematical constraints.
We presented CAP System that is domain independent, easily extensible, and
could be used in various types of context-aware personal services, based on
our semantic technology adoption strategy.
Finally, we illustrated how CAP was used to support the implementation of a
demo context-aware personal service with acceptable performance and
scalability.
Future Work
privacy and security feature
rule mining for more intelligent services
test and evaluate our system using actual hardware devices.
Copyright 2008 by CEBT
Semantic Tech & Context - 12
Current CAP System – Server & Client
Altitude
Sensor
Humidity
WWW
Real World Environment
Temperature
Pressure
Sensor
Sensor
…
Location
Device (Client)
Device (Client)
Device (Client)
Device (Client)
Virtual Sensor
Server
Interface
GUI
Interface
Request Processor
Context Manager
Context
Rules
Vocabulary Ontology
Privacy Manager
Privacy
Policy
Service Trigger
Service Repository
Legacy System
Service
Processor
Context
Data
Service
Rules
Context-aware Service
Description Language (CASDL)
Log Manager
Log
Service
Description
Copyright 2008 by CEBT
Semantic Tech & Context - 13
Thank You
Thank You
감사합니다
Copyright 2008 by CEBT
Semantic Tech & Context - 14