An Architecture for Adaptive Content Extraction in Wireless Networks

1
An Architecture for Adaptive Content Extraction
in Wireless Networks

• Phil West
• Greg Foster
• Peter Clayton

Submitted to the South African Telecommunications
Networks Applications Conference (SATNAC) 2005
2
Overview

• Scenario
• Proxy Server
• Spatial Representation
• Tag Extraction
• Personalization
• Monitoring Network Conditions
• Adaptive Extraction
• Architecture Putting it all together
• Current Work

3
Scenario
4
Proxy Server

• Crunch 2.0
• Open-source, Java project
• Thread-based proxy server
• Extensible
• Plugins using reflection
• Each plugin receives a DOM tree representation of
  the content
• Contains an extraction module
• Advertisements
• Plain text from HTML For visually impaired
  users
• Link lists
• Empty tables

5
Spatial Representation

• Process
• Parse HTML into DOM tree
• Render the nodes using ICEBrowser SDK
• Extract boundbox for each node
• Generate image based on coordinates
• Scaled to fit on PDA screen
• Create HMTL image map to allow navigation

6
Spatial Representation

• Issues
• Script
• Algorithm complexity
• Depth of trees
• Size of nodes
• Other approaches
• Use for special cases
• Structured HTML/XML

7
Tag Extraction

• Extract rich content
• Simply replace tags with links
• Issues
• May lose context extracted content displayed in
  a new window

8
Personalization

• Server-side
• SQL database stores session preferences for each
  user
• ipaddress settingsspatial,video,sound,etc
• Settings for each request them obtained from DB
• Client-side
• Default settings stored locally
• Browser Helper Object (BHO) as a plug-in for
  Internet Explorer
• BHO listens for navigate2 events
• Appends settings to request if
• (first request of session settings have been
  altered)

9
Monitoring Network Conditions

• WinPcap Windows Packet Capture Library
• Used to access network statistics
• Implemented a class in C
• Java Wrapper Class Exposes Native Methods
• Provides
• Bytes per second
• Packets per second
• Also Track
• Number of connections (in a defined time period)

10
Adaptive Extraction

• Load Testing
• Microsoft Application Centre Test
• Proxy Sniffer
• Created test web pages
• Load test these pages, accessed through the proxy
  server

• Content extracted at increasing levels of network
  usage
• 1st Video and sound
• 2nd Script
• 3rd Images
• Increase the number of possible simultaneous
  connections

11
Putting it all Together

• Adaptive Content Extraction Architecture

12
Putting it all Together

• Provides
• Mark-up language, content extraction system for
  use with wireless networks
• Users have the ability to select only the content
  they desire
• Implications for pay-for-data billing schemes
• Functionality is provided to allow users to
  personalize their extraction settings
• Algorithm for automatically adapting extraction
  settings according to varying network conditions
• Preliminary results show this allows more users
  to access the network without losing performance
• 2 real world implementations

13
Current Work

• Adapting the approach for use with slide show
  presentations
• Will be presented at the 4th World Conference on
  mLearning
• Functionality to annotate content
• Delta correction
• Usability study
• Investigating alternative HTML renderers
• For use with the spatial representation algorithm
• More in-depth testing of the adaptive system

14
(No Transcript)