Introduction - Department of Computer Science and Engineering

Download Report

Transcript Introduction - Department of Computer Science and Engineering

Second semester FYP Presentation 2001~2002
LYU0101
Wireless Digital Information
System
Supervisor Prof. Michael Lyu
Yeung Kam Wah
Lam Yee Gordon
Overview





Introduction
Design Direction
System Architecture
Work Done
Possible Improvement
Introduction

PDA become more popular


High Bandwidth Wireless Technology



Computation power and Memory is
increasing
BlueTooth, Wireless LAN
3G
Demand for provide content to PDA

Provide location related information
Wireless Digital Information System

To build a generic system that allow
company provide their content to user




Support different application
Single client for different server
Search function
Support multimedia content
Wireless Digital Information System

Usage




Shopping guild in a Shopping Mall
Information search in an Exhibition
Showing vacancy in a Car Park
Digital video Library client
Design Direction
Design Direction


What’s our goal? (in previous slide)
Main Issues we need to concern:



Data (format, storage, types)
Search Engine (vector,probabilistic model,
indexing, ranking, relevance feedback)
Platform (hand-held device)
Concern 1:Data

Use XML as the primary data format


To give our user a standard, feasible and
extensible way to design the data structure
separate the concern on the data and
presentation of information


Use XSL to change XML into information
changes in module or client platform will not
affect the data
Concern 1:Data

Storing XML in Database


Database provides us many useful services, e.g.
building different kinds of indexing on data,
concurrency control, recoverability of data, etc.
Break down XML file into tree nodes and store
as tuples (parent id, node id, type, data)


easy for locate and update the data in XML
reform the XML on need
Concern 2: Searching

Use Vector Model



assign weight to index terms in queries
and in XML file (from tf factor, termfrequency and idf factor, inverse document
frequency)
compute degree of similarity between each
XML file and user query
give ordering of XML file
Concern 2: Searching

A full inverted index is built.


have a better ranking, e.g. search for “林
怡”, I prefer to have “林怡” in search
result, rather than “林保怡”.
can help in searching a whole phase.
Concern 2: Searching



Problem: client may need only ten
search results at a time, I can’t send all
the result to PDA client at a time !!! (in
decreasing of the similarity, just like
have different result page in google)
NN-search Methods (R-tree, SS-tree, SRtree, X-tree, MVP-tree, A-tree, VA-file)
We choose VA-file
Concern 2: Searching

Reason for choosing VA-file:

performs much better than the others in
high-dimension space(as number of index
term may be over thousand)



IO page, number of Vector calculation, CPU
time in search
CPU time on construction
Characteristics of VA-file:

no MBRs, no Tree, use approximation
linear search, just build a small filter file
Concern 3: Client Platform


We choose Pocket PC as our client
platform
Reason:

Pocket PC have better support in



Internet
Multi-Media
Performance

processing power of Pocket PC is higher than
Palm OS nowadays
System Architecture
System Architecture
Pre-Process
System
XML
Database
Index
Search Engine
request
PDA client
Network
TCP/IP, HTTP
XML/HTML
Media server
server
Query & Result
XLS
System Architecture

Server side



XML Data is processed and stored in the
database
A set of XSL is used to format the output
Client side


Displace information
Send user request to server
Server Side

Four main components at Server Side




XML Pre-processor
Search Engine
XSL transformer
TCP Server that Communicate with PDA client




incoming: Keyword search, Map search result,
<XSL,XML>
outgoing: HTML, Video xml definition (probably a very
simple SMIL), Map xml definition
Web server (images)
MMS server (videos)
XML Pre-processor
XML Data
Struture + Text
Strutural
parsing
tuples
Text
Accents,
spacing, etc
Lexical
Analysis
Building Full
Inverted index
Vector and
VA-file Builder
Vectors and
VA-file
Relational Database
The PDA Client

The four main components




Control component
Main window
Video window
Image window
The PDA Client

Main Window




HTML Viewer
Supported by system provided DLL
Interface for information displaying
User event will pass to control component
to handle
The PDA Client

Video/Audio window



Supported by system provided COM Object
Display video to user
Image Window



Using to display image
Points highlight feature
Using as query input interface, user can
select a rectangular area as query input
The PDA Client

Why using HTML viewer as main interface



Provide feasibility to different servers using their
own interface
HTML is widely used and designed for display
information.
Why using system provide components


Simplify the program and slim the program size
These component is optimized
Work Done and Possible
Improvement
Work Done

2001 July ~ Aug




Study different wireless network
Study different PDA platform
Build a trial application on Pocket PC
2001 Sep ~ Dec


Design the application and system
architecture
Build the Packet PC client
Work Done

2002 Jan ~ April



Study the searching algorithm
Build the server
Continue implement the Pocket PC client
Possible Improvement

At Server Side:



relevance feedback (improvement on query)
caching (better performance)
At Client side:

enhancement on the Main window, Image
window and Video window