Web Data Management Panel Presentation WITS

1
Web Data ManagementPanel Presentation WITS 97

• Joachim Hammer
• University of Florida
• December 14, 1997

2
World Wide Web

• Easy to use interface (pages and links)
• Ubiquitous
• Excellent storefront
• Lots of valuable information
• Irregular structure (semistructured)
• Highly dynamic
• Difficult searching/navigation
• Limited querying
• Human is query processor

3
Web Management Issues Research

• Information browsing through the Web
• Effective, simple to program GUI
• MOBIE (Stanford University)
• Dealing with existing (static) Web data
• Cant query static Web pages
• TSIMMIS Data Extraction (Stanford University)
• Information discovery
• Too much quantity, too little quality
• WebMining (University of Florida)

4
MOBIE

• Formats and displays data objects as a web of
  hypertext documents
• Traverse hyperlinks to explore nested structure
  and contents
• Based on HTTP and HTML
• Provides simple, world-wide access to information
  servers
• New way of exploring databases
• Much like readers explore contents of a book

5
Raw Data Object
ltcollection, b1, a1, ...gt b1 ltbook, t, agt
t lttitle, Database and ...gt a
ltauthor, Jeff Ullmangt a1 ltarticle, v, w,
xgt v lttitle, Mediators in ...gt
w ltauthor-list, ...gt x ...
...
. . .
6
Formatted - Hyperlinked
collection book title Database and
... author Jeff Ullman article
title Mediators in the ... author Gio
Wiederhold
7
Data Extraction Querying

• Configurable parser (Python)
• Declarative description of HTML source
• Location of data on page
• How to package data into result object
• Regular expression-like syntax
• Human intelligence rather than A.I.
• Returns data as OEM (Object Exchange Model)
  objects
• TSIMMIS interchange format

8
Approach

• Extract data from Web page(s)
• On demand
• Periodic/on update
• Use wrapper/DBMS as query processor

Wrapper
Query/ Result
World Wide Web
Extractor
or
Persistent Storage
Query/ Result
Specification
9
Evaluation

• Better than
• Writing programs
• YACC, PERL, etc.
• Want to do even better
• GUI tool to simplify the generation of extractor
  specification

10
Information Discovery

• Improve existing search engines
• Efficient crawling techniques (reduce data
  shipping)
• Quality ranking of pages
• Apply data mining techniques to categorize web
  pages
• e.g., clustering algorithms, proximity
• Inferencing of knowledge
• Making connections among entities

11
Goals of Web Management

• Putting vast amounts of previously unavailable
  data on the Web
• Digital libraries, long distance learning,
  research, etc.
• Data managed by DBMS
• Leverage existing DBMS technology
• New tools for managing semistructured data
• Support of electronic storefronts
• Dynamic creation of customizable Web pages