Analysis for the W3C Glossary and Dictionary Project

Overview

The Glossary project can be divided into three main parts:

a User Interface
a back-end
a data model

User Interface

Users

Editor
Translator
Reader

Back-end

Will handle 2 types of action:

updates (add, modify, delete)
queries

This is a summary of the use cases defined so far in the Glossary requirements and during the March 7 2003 Glossary meeting:

Update use cases

Use cases	Editor (ED)	Translator (TR)
add term	ED
modify term	ED
delete term	ED
add definition	ED
modify definition	ED
delete definition	ED
add translation		TR
modify translation		TR
delete translation		TR

Query use cases

Use cases	Editor (ED)	Translator (TR)	Reader (RE)
comparison	ED	TR	RE
discovery	ED	TR	RE
derivation	ED
policy making		TR	RE
illustration?
speech synthesis			RE
cross-references	ED

Currently experimenting with the following tools to manipulate RDF:

Data model

as described in the March 7 2003 Glossary meeting notes.

term / concept
definition
context
contributors
date
translations (term, definition)
comments
synonyms
other data models - ISO

Dominique Hazaël-Massieux suggested (for the time being) that one glossary be mapped into one RDF schema.
For instance, if we create an RDF schema for the SpecGL glossary (http://example.org/specGL/20030210), such a schema would contain an entry for each term as shown below:

<rdf:Description rdf:about="http://example.org/specGL/20030210#discretionaryChoices">
  <rdfs:label xml:lang="en">discretionary choices</rdfs:label>
  <rdfs:comment>a value or behavior may be chosen from a well-defined enumerated set of two or more possibilities</rdfs:comment>
  <rdfs:isDefinedBy rdf:resource="http://www.w3.org/TR/2003/WD-qaframe-spec-20030210/"/>
  ...
</rdf:Description>

Those RDF schemas will be used to get the glossary database started.

Scraping system

Documents written in XHTML

Extraction rules

In order to work properly, the document should be pubrules compliant.

markup
- ignore definition lists that immediately follow
  - <h2>References</h2>
  - <h2>Change history</h2>
  - <h2>Abbreviations</h2>
  - <h2>Acronyms</h2>
  - ...
class conventions:
- ignore definition lists inside a div element that has a class attribute with value head
- keep definition lists that have a class attribute with value glossary

Input, suggestions welcomed.

XSLT stylesheets

html-glossary-to-rdf-schema.xsl: extracts the information: metadata about the document (relies on trdoc-data.xsl) and glossary data
rdfize.xsl: formats the data
glossary-util.xsl: miscellaneous templates
Util.xsl: XSLT templates to manipulate strings

RDF schemas

TestGL.rdf: glossary of QA Framework: Test Guidelines
Intro.rdf: glossary of QA Framework: Introduction
SpecGL.rdf: glossary of QA Framework: Specification Guidelines
OpsGL.rdf: glossary of QA Framework: Operational Guidelines

Documents written in XMLSpec

XSLT stylesheets

xmlspec-glossary-to-rdf-schema.xsl: extracts the information: metadata about the document (relies on xmlspec-trdoc-data.xsl) and glossary data
rdfize.xsl: formats the data
glossary-util.xsl: miscellaneous templates
Util.xsl: XSLT templates to manipulate strings
xmlspec-trdoc-data.xsl

xmlspec-trdoc-data.xsl is the first attempt at translating the templates defined in trdoc-data.xsl so that they work with XMLSpec documents.

xmlspec-trdoc-data.xsl issues

the dependence between the template getDate and getSubtitleStatus has been reversed: getSubtitleStatus now relies on getDate, whereas it was the other way around in trdoc-data.xsl

xmlspec-trdoc-data.xsl "to do" list

Some templates defined in trdoc-data.xsl have not yet been translated for various reasons:

getEditorsSectionName: not sure that it will be useful
getActivity: have to figure out how the original template works
formatActivityHP & formatActivityStatement: see above
getFeedbackDueDate: have to figure out how the original template works

Preliminary results

Below are the results of applying the currently implemented templates (Results column) and the xmlspec-glossary-to-rdf-schema.xsl stylesheet (RDF Schemas column)to the XMLSpec version of some documents:

HTML	XMLSPec	Results	RDF Schema
XPath 1.0 Recommendation	REC-xpath-19991116.xml	xpath.txt	xpath.rdf
XML 1.0 Recommendation	REC-xml-20001006.xml	xml.txt	xml.rdf
XLink 1.0 Proposed Recommendation	Overview.xml	xlink.txt	xlink.rdf
DOM Level 3 XPath 1.0 Candidate Recommendation	xml-source.xml	dom_level_3_xpath.txt	dom_level_3_xpath.rdf
Character Model 1.0 Working Draft	Overview.xml	charmod.txt	charmod.rdf
XSL 1.0 Working Draft	xslspec.xml	xsl.txt	xsl.rdf
WSDL 1.2 Working Draft	wsdl12.xml	wsdl.txt	wsdl.rdf

Project roadmap and schedule

NB: the schedule below is provided for information only

testing of the tools listed above to determine the most suitable one for the project and for me: 2 to 4 weeks (~ end of April)
implementation of the use cases defined so far, i.e. translate the use cases into RDF query language queries (if possible): at least 2 months (~ end of June)
design of the user interface: 1 month (~ end of July)
test phase: remaining time until the end of my internship (August 29 2003)

On-going work

extraction rules enhancement
definition of the data model and related RDF schema