Automatic RDF Metadata Generation for
Resource Discovery
Charlotte Jenkins, Mike Jackson, Peter Burden, Jon Wallis
School of Computing & IT
University of Wolverhampton
Wulfruna Street, Wolverhampton
WV1 1SB, UK
Abstract
Automatic metadata generation may provide a solution to the problem of inconsistent,
unreliable metadata describing resources on the Web. The Resource Description Framework
(RDF [10]) provides a domain-neutral foundation on which extensible element sets can be
defined and expressed in a standard notation. This paper describes how an automatic
classifier, that classifies HTML documents according to Dewey Decimal Classification
(DDC [8]), can be used to extract context sensitive metadata which is then represented using
RDF. The process of automatic classification is described and an appropriate metadata
element set is identified comprising those elements that can be extracted during
classification. An RDF data model and an RDF schema are defined representing the element
set and the classifier is configured to output the elements in RDF syntax according to the
defined schema.
Keywords:
RDF, metadata, classification.
1. Introduction
A major problem facing tools for information resource discovery on the Web is the lack of a
mechanism for resource description within the Web's architecture. There are now said to be
in excess of 320 million individually accessible objects on the Web [5]. There is no one
accurate, reliable, up-to-date, comprehensive method of finding out what each one of these
objects is, what type of resource it is, what the subject matter is and so on, without accessing
and analysing each one individually. This is a problem, not only for resource discovery, but
also for content rating where illicit material is concerned. The World Wide Web Consortium
(W3C [13]) has introduced the Resource Description Framework (RDF), in an attempt to
produce a standard language for machine-understandable descriptions of resources on the
Web. RDF is intended to support resource descriptions for resource discovery and also for
rights management, privacy preferences, content ratings (PICS [9]), evaluation and
classification. RDF is seen as the framework for producing a Web of trust where the content
of each individually accessible object is well described in a format that is extensible yet
universally understood. RDF may enable search engines and other tools for resource
discovery to exchange and share metadata. This paper is concerned with the automatic
generation of metadata in RDF format for use in describing HTML documents for the
purposes of resource discovery.
Page 1 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

Various attempts have been made to introduce embedded metadata into HTML documents,
most notably through the use of the HTML META tag and embedded Dublin Core [12]. It is
also now possible to include an embedded RDF description of a document. The problem with
such techniques is that they are not compulsory so many authors still choose not to include
meta information. M. Marchiori, in his paper entitled The Limits of Web Metadata and
Beyond [7], addresses this issue by proposing a scheme that involves back-propagating meta
information from pages with known metadata to those that are linked from it. An alternative
method of automatically generating metadata is to use an automatic classifier. The automatic
classifier described in this paper works by comparing terms found within documents with
manually defined clusters of terms representing the nodes of a classification hierarchy
(DDC). This process results in the identification of other useful metadata such as the
document title, keywords, abstract and word count in addition to the classification
classmarks. An RDF schema has been defined for representing this metadata and the process
by which it is extracted and represented in RDF is described.
2. Automatic Classification
The automatic classifier [3] described below has been designed and developed for use as an
automated component of a distributed automated search engine. The use of automatic
classification within an automated search engine is quite unusual - commonly automated
search engines (such as AltaVista) are huge indexes and classified tools (such as Yahoo and
Galaxy) require some degree of manual intervention, typically in specifying the classification
category and other such meta information. It has been observed [6] that, classified tools,
although often hopelessly incomplete and out-of-date because of the lack of automation, are
less likely to inundate users with irrelevant information. Automatic resource discovery
combined with automatic full text indexing is faster and more comprehensive than manual
classification but much less accurate. It is hoped that the use of automatic classification will
combine the advantages of both approaches resulting in an accurate, comprehensive, up-to-
date, well classified, automated search engine. Documents sharing the same subject matter
will be automatically clustered together under the same classification classmarks and
therefore will be retrieved together more easily.
The automatic classifier classifies documents according to DDC. DDC is considered
appropriate because it is a universal classification scheme covering all subject areas and
geographically global information. It is familiar to anyone accustomed to using a library and
has multilingual scope. The hierarchical nature enables the users of a search engine to refine
their search from rough classifications to increasingly more accurate ones.
The automatic classifier is an object oriented system, written in Java, that retrieves HTML
documents from given URLs, analyses the contents and assigns appropriate DDC
classmarks. A hierarchy of Java classes is used to model the DDC classification hierarchy.
Documents are filtered through this hierarchy according to which class representatives
(manually defined terms representing each DDC class) best match the document's contents.
Each term found within the document is given an associated weight which is greater if the
term is found in the title or a heading element. Terms found within the keywords or
description elements of existing META tags are also stored with significant associated
weight. Terms also acquire more weight the more often they occur. These weighted terms are
then compared with the manually defined terms representing DDC classes. Initially the
Page 2 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

document is compared with the top ten DDC classes shown in Figure 1.
000 Generalities
100 Philosophy, paranormal phenomena, psychology
200 Religion
300 Social sciences
400 Language
500 Natural sciences and mathematics
600 Technology (Applied sciences)
700 The arts, Fine and decorative arts
800 Literature (Belles-lettres) and rhetoric
900 Geography, history, and auxiliary disciplines
Figure 1. The ten DDC classes
If a significant match is found between the document and a DDC class, the document is then
compared with subclasses of that DDC class. This comparison process continues recursively
through the hierarchy until significant matches with leaf nodes are found, the classmarks of
which are assigned to the document.
To illustrate this process more clearly, figure 2 shows a document object which comprises:
an accession number that is used to uniquely identify the document
a series of keyword objects, each one representing a term found within the document,
with an associated weight depending on where it was found within the document and
how frequently it occurs (note, ALL found terms are stored in this manner)
a series of classmark objects, each one comprising the actual classmark together with a
textual label e.g. 303.483 Development of science and technology. Appropriate
classmark objects are assigned here when the keywords match significantly with the
keywords of DDC objects that have no subclasses (see figure 3) i.e. leaf nodes in the
hierarchy.
Figure 2. A document object comprising a series of keyword objects and a series of
classmark objects
Page 3 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

Figure 3 shows a DDC object which comprises:
a series of keyword objects, identical in structure to the document keywords but
comprising manually defined terms representing this particular DDC class
a series of subclasses - each of which is itself a DDC class representing the next layer
of the hierarchy beneath this class. Leaf nodes obviously have no subclasses
a classmark object defining and uniquely identifying this class. If the keywords of this
class match significantly with the keywords of a document object and there are no
subclasses, this classmark object is assigned to the document.
Figure 3. A DDC object comprising a series of keyword objects, a series of DDC objects
representing its subclasses and a classmark object
The hierarchical nature of the DDC classes enables ambiguous terms to be concealed and
considered in context lower down the class hierarchy. The class representatives at the top of
the hierarchy contain a broader range of terms than those nearer the bottom which are more
detailed and more specific.
Measures of similarity between the document and DDC class representatives are calculated
using the Dice Coefficient [11]:
Figure 4. The Dice Coefficient
Each time a word in the document matches a word in the DDC class representative, the two
Page 4 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

associated weights are added to a total score (x intersection y). This score is then divided by
the sum of the number of keywords in the document and the number of keywords in the class
representative (x union y) and the result is multiplied by 2. Any result greater than 0.5 is
considered significant and the document will proceed to be compared with any subclasses or
be assigned the classmark if there are no subclasses. If the score is not significant, the
comparison process will proceed no further through this branch of the hierarchy.
The comparison process may proceed through several unrelated branches of the hierarchy for
as long as significant matches are found. In a Web library multiple classifications are
appropriate - the same book can appear on several different shelves.
2.1 Metadata Elements
The classification process results in the production of a series of classmarks appropriate to
describe a particular document. However, the process can easily be used to pull out various
other metadata elements. During the parsing of the document, terms found in the title
element are singled out as being important, these can easily be extracted as can those terms
which match those found in the class representatives of appropriate DDC leaf nodes i.e.
significant keywords. Keywords and descriptions found in existing META tags can be
extracted. Other useful metadata that is easily accessible is shown in Figure 5. This element
set is based on those metadata elements that can easily be obtained from the process of
automatic classification. These elements are particularly suited to the domain of the
automated search engine.
Element Description Purpose
1. Unique
accession
number
Number assigned by the system. Uniquely
identifies the
resource.
2. Title Taken from the HTML <TITLE> element. Usually helps in
discerning the
subject matter.
3.
URL
*
The URL given to the system, used to extract the
document for classification.
Indicates the
location of the
document.
4. Abstract Either the first 25 words found in the body of the
page, or, if present, taken from the Description
META tag. (A much more sophisticated
abstracting technique could be used here in future
implementations).
Provides further
clues about the
subject matter.
5. Keywords Terms found within the document that match
terms found within the class representatives of
DDC classes found to be appropriate.
Indicate key
issues/topics.
6. Classmarks DDC classmarks found to be appropriate as a
consequence of the classification process.
Indicate subject
area(s).
7. Word count The number of words found on the page,
including the title.
Indicates extent,
detail, download
time.
Page 5 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

* The classifier only handles individual HTML documents so the URL, not URI, is appropriate. The URL is not
used as an identifier within the search engine because it is possible for the same page to have more than one
URL; this is one of the causes of repetitions in automated search engine results.
Figure 5. An appropriate metadata element set - The " Wolverhampton Core "
8. Classification
date
The system date when the classification took
place (GMT or BST)
Indicates currency
of the metadata.
9. Last modified
date when
classified
Taken from the HTTP Last-modified header.
(Gives Not known if equal to the "epoch" - 1
st
January 1970)
Indicates currency
of the information.
It is thought that these elements (Wolverhampton Core) are sufficient to uniquely identify the
document, state where it can be found, provide a good indication of the subject matter and of
how current both the actual information and its metadata are.
The most well known and well used metadata element set for resource discovery is Dublin
Core [12]. Compliance with a recognised standard is advisable because it encourages
interoperability and consistency between applications. Dublin Core has evolved from the
Digital Library community and consequently not all of its elements are as well suited to the
automated search engine domain as those defined in figure 5. There is, however a significant
overlap and none of the Dublin Core elements are compulsory. RDF enables developers to
tailor an element set to suit their application while still reusing appropriate standard elements
defined elsewhere (see section 3).
Figure 6 compares the fifteen elements of Dublin Core with the elements defined in figure 5.

Dublin Core Elements Equivalent Wolverhampton Core Elements
1.
Title Title
2.
Creator -
3.
Subject Keywords + Classmarks
4.
Description Abstract
5.
Publisher -
6.
Contributor -
7.
Date Last modified when classified
8.
Type -
9.
Format -
10.
Identifier Accession number + URL
11.
Source -
12.
Language -
13.
Relation -
14.
Coverage -
Page 6 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

Figure 6. Comparison between Dublin Core and the Wolverhampton Core element sets.
15.
Rights -
16.
- Date Classified
17.
- Word count
It can be observed that most of the Wolverhampton Core elements have a Dublin Core
equivalent. The implications of this comparison are discussed again in the later section (3.2)
on RDF schema definition. It is thought that the specified Wolverhampton Core elements
represent an appropriate subset of Dublin Core (with one or two additions) that is suited to
the requirements of an automated search engine.
Once the necessary metadata elements have been identified they can then be represented in
RDF.
3. Resource Description Framework (RDF)
Three things are required in order to generate RDF statements about a resource: a data
model, a schema and the actual representation in XML (eXtensible Markup Language [2])
syntax. Several RDF schemas might actually be involved; schemas are required for the
interpretation of RDF statements. The following three subsections explain how the metadata
elements shown in figure 5 can be represented by an RDF data model, defined using an RDF
schema and, most importantly, automatically generated in RDF/XML syntax.
3.1 RDF Data Model
The RDF data model is expressed using directed labelled graphs (or "nodes and arcs"
diagrams) which identify the properties and property values associated with a resource as
shown in figure 7. (This notation is taken from the RDF Model and Syntax Specification
[4]).

e.g.
Figure 7. Data Model notation showing an RDF statement; a resource, a named property
and the value of that property.
Page 7 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

In RDF a resource may be a simple Web page, part of a simple Web page, a collection of
pages or a whole Web site. The automatic metadata generator described in this paper
generates descriptions of individual HTML pages.
Figure 8 shows how the element set in Figure 5 would be represented for the HTML page at
http://www.scit.wlv.ac.uk/index.html
Figure 8. An RDF Data Model using the elements from Figure 5.
The model shows two RDF containers - one a bag of keywords and the other a sequence of
classmarks. The classification process will usually result in the identification of several
keywords within the document but the order in which they are presented is insignificant so a
bag is appropriate. A better method of representing the keywords would be to use a Set
Page 8 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

where no duplicates would be permitted, however, RDF does not define a Set because there
is no defined enforcement mechanism in the event of violation. The classmarks are ordered
by the classifier according to which scored the highest measure of similarity and so these are
represented as an ordered sequence. The classmarks would be better represented by an
ordered collection class where no duplicates would be allowed. Further work layered on the
RDF core may define such enforcement mechanisms.
3.2 RDF Schema Definition
Once the appropriate properties have been identified a schema must be created, or existing
schemas must be identified, where these properties are defined. Schemas provide the RDF
type system and enable applications to interpret RDF statements (see next subsection). The
properties could be expressed using appropriate existing vocabularies, in which case it is not
necessary to define a new schema - existing schemas can be referenced from within the
RDF/XML syntax. It is possible to reference as many different schemas as required, mixing
and merging different vocabularies. Schemas are referenced using the namespace mechanism
from within the RDF syntax (see next subsection).
The definition of new schemas enables developers to specify properties best suited to their
particular application. Schemas can define properties that are sub-properties of those defined
elsewhere in existing schemas. This feature has been utilised in the Wolverhampton Core
schema definition shown in appendix A. A property has been defined for each element
identified in figure 5. Those elements that Wolverhampton Core has in common with Dublin
Core (as shown in figure 6) have been defined as sub-properties of the appropriate Dublin
Core properties. (The Dublin Core properties are based on those shown in the example
Dublin Core schema in the RDF Schema Specification [1]. Note, this is not the authoritative
Dublin Core schema which will be made available by the Dublin Core Initiative). This
approach has been adopted so that Wolverhampton Core properties have specialisation
relationships with Dublin Core properties and retain some implementation freedom. It would
have been possible to use the Dublin Core properties directly in the automatically generated
RDF syntax (see next subsection) but it is very important that the automated search engine is
clear about the particular implementation of these properties. For example, both keywords
and classmarks could be expressed as Dublin Core Subject properties (see figure 6) but the
search engine needs to be able to differentiate between keywords and classmarks. Two
Wolverhampton Core properties are defined, representing the keywords and classmarks
independently, both of which are defined as sub-properties of the Dublin Core Subject
property. Creating specialisation relationships in this manner will enable applications capable
of processing Dublin Core to, at least partially, interpret Wolverhampton Core thereby
encouraging both interoperability and extensibility.
The schema for defining the Wolverhampton Core element set can be found in Appendix A.
(The schema specification language in which this schema is written is defined in the RDF
Schema Specification [1]). This is a very simple definition of the properties required to
represent the elements identified in figure 5. Future implementations could define new
classes and declare constraints on the properties.
3.3 RDF Syntax
The following shows the RDF representation of the data model shown in figure 8. Appendix
B shows automatically generated RDF for a series of test URLs. (The RDF/XML syntax
Page 9 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

used here is described in The RDF Model and Syntax Specification [4].)
<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about="http://www.scit.wlv.ac.uk/">
<wc:Accession_no>583295</wc:Accession_no>
<wc:Title>SCIT WWW Server</wc:Title>
<wc:Abstract>
School of Computing and Information Technology
WWW server General Information University of
Wolverhampton School of Computing and Information
Technology home page Wolverhampton and surrounding
areas
</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>computer</rdf:li>
<rdf:li>computing</rdf:li>
<rdf:li>database</rdf:li>
<rdf:li>databases</rdf:li>
<rdf:li>server</rdf:li>
<rdf:li>search</rdf:li>
<rdf:li>searching</rdf:li>
<rdf:li>directory</rdf:li>
<rdf:li>university</rdf:li>
<rdf:li>school</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Bag>
<rdf:li>005.7 Data in computer systems</rdf:li>
<rdf:li>370 Education</rdf:li>
</rdf:Bag>
</wc:Classmark>
<wc:Word_count>216</wc:Word_count>
<wc:Last_modified>20/11/1997</wc:Last_modified>
<wc:Classification_date>11/09/1998</wc:Classification_date>
</rdf:Description>
</rdf:RDF>
Note that there are two XML namespace definitions (xmlns) at the top of this piece of RDF.
The first one identifies the location of the RDF syntax specification and the second one
identifies the location of the Wolverhampton Core (wc) schema where the property types
specified within this RDF description are defined. This wc schema is shown in Appendix A.
W3C and the Dublin Core Initiative recommend the use of ISO 8601 Date format. This has
not been implemented in this instance because the automatic metadata generator is to be
deployed as part of a UK search engine where dates will be required in UK format.
If the classification process should fail, i.e. no significant measures of similarity are found,
other elements such as the title, abstract, word count and dates should still be identified.
4. Conclusions
Although it is envisaged that the editing tools of the future will encourage the inclusion of
RDF meta information, the current situation, where some authors choose not to include any
metadata, is likely to continue to some extent. It is very difficult to automate resource
description but it would be impossible to describe everything on the Web manually.
Automatic metadata generation would appear to be an essential pre-requisite for widespread
deployment of RDF based applications. The Web of trust must attempt to be comprehensive
Page 10 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

because a Web that is partially trust worthy offers little advantage over one that cannot be
trusted at all, especially where content rating is concerned.
The automatic metadata generator described in this paper enables an RDF description to be
associated with any HTML page, regardless of when it was created and by which editing
tool. RDF has enabled the specification of a metadata element set that is tailored to suit an
automated search engine but strongly related to a standard, digital library element set, Dublin
Core. The ability to create specialisation relationships with appropriate Dublin Core
properties increases the potential for interoperability - any application capable of processing
Dublin Core will be capable of processing most of the defined Wolverhampton Core
properties because they are defined as sub-properties of Dublin Core properties. Such
interoperability will encourage information sharing which will improve comprehensive Web
coverage; if search engines can process the same standard syntax, they will be able to
exchange metadata and integrate their results. Some subject-specific classified directories are
known to be attempting to share information through the use of RDF already; information
sharing between automated search engines has even greater potential.
References
1. D. Brickley, R. Guha, A. Layman, Resource Description Framework (RDF) Schema
Specification, http://www.w3.org/TR/WD-rdf-schema (Working Draft - October 1998)
2. D. Connolly, J. Bosak, Extensible Markup Language (XML),
http://www.w3.org/XML/, (October 1998)
3. C. Jenkins, M. Jackson, P. Burden, J. Wallis, Automatic Classification of Web
Resources using Java and Dewey Decimal Classification, Computer Networks and
ISDN Systems, Volume 30 646-648 (1998)
4. O. Lassila, R. Swick, Resource Description Framework (RDF) Model and Syntax
Specification, http://www.w3.org/TR/WD-rdf-syntax (Working Draft - October 1998)
5. S. Lawrence, C. L. Giles, Searching the World Wide Web, SCIENCE, Volume 280
(April 1998)
6. L. Lindop, M. Sriskandarajah, M. Williams, M. Bracken, M. Cadden, A. Dabbs, W.
Gallagher, Catching Sites, PC Magazine 6 (2) (February 1997)
7. M. Marchiori, The Limits of Web metadata and beyond, Computer Networks and
ISDN Systems, Volume 30 1-9, (1998)
8. OCLC Forest Press, Dewey Decimal System Home Page,
http://www.oclc.org/oclc/fp/index.htm (October 1998)
9. P. Resnick, Platform for Internet Content Selection (PICS), http://www.w3.org/PICS/

(January 1998)
10. R. Swick, E. Miller, B. Schloss, D. Singer, Resource Description Framework (RDF),
http://www.w3.org/RDF/ (October 1998)
11. R. C. J. van Rijsbergen, Information Retrieval: Second Edition, Chapter 3,
http://www.dcs.glasgow.ac.uk/Keith/Chapter.3/Ch.3.html Butterworths, ISBN 0-408-
10775-8, 1981
12. S. Weibel, E. Miller, Dublin Core Metadata, http://purl.oclc.org/metadata/dublin_core/
(November 1998)
13. The World Wide Web Consortium, http://www.w3.org/ (October 1998)
Page 11 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

Appendix A
Below is the RDF schema for the Wolverhampton Core (wc) element set referred to in figures 5 and 8.
Note that the URL is not specified as a separate property because it is always noted in the
<rdf:Description about="http://...">
statement.
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:rdfs="http://www.w3.org/TR/WD-rdf-schema#">
<rdf:Description ID="Accession_no">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:subPropertyOf resource="http://purl.org/metadata/dublin_core#Identifier"/>
<rdfs:label>Accession_no</rdfs:label>
<rdfs:comment>A unique number assigned by the automatic classifier
that uniquely identifies this resource.</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Title">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:subPropertyOf resource="http://purl.org/metadata/dublin_core#Title"/>
<rdfs:label>Title</rdfs:label>
<rdfs:comment>The title of the resource taken from the HTML TITLE element.
</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Abstract">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:subPropertyOf resource="http://purl.org/metadata/dublin_core#Description"/>
<rdfs:label>Abstract</rdfs:label>
<rdfs:comment>This is the first 25 words taken from the BODY of the HTML
page, or, if present, text taken from the description HTML META tag.
</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Keyword">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:subPropertyOf resource="http://purl.org/metadata/dublin_core#Subject"/>
<rdfs:label>Keyword</rdfs:label>
<rdfs:comment> This is a keyword from the document that matched a keyword
in an appropriate DDC class representative. A number of keywords will
normally appear in an RDF Bag container.</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Classmark">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:subPropertyOf resource="http://purl.org/metadata/dublin_core#Subject"/>
<rdfs:label>Classmark</rdfs:label>
<rdfs:comment>This is a DDC classmark that has been assigned to the document
as a result of the automatic classification process. Often two appropriate
classmarks will be shown in an RDF sequence - the highest scoring one
appearing first.</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Word_count">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:label>Word_count</rdfs:label>
<rdfs:comment>This is the number of individual words found in the
resource.</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Classification_date">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:label>Classification_date</rdfs:label>
<rdfs:comment>The date on which the resource was classified.</rdfs:comment>
</rdf:Description>
<rdf:Description ID="Last_modified">
<rdf:type rdf:resource="http://www.w3.org/TR/WD-rdf-syntax#Property"/>
<rdfs:subPropertyOf resource="http://purl.org/metadata/dublin_core#Date"/>
<rdfs:label>Last_modified</rdfs:label>
<rdfs:comment>The date on which the resource was last modified
when it was classified.</rdfs:comment>
</rdf:Description>
Page 12 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

</rdf:RDF>
Appendix B
The following RDF descriptions have been automatically generated. The automatic metadata generator
is a Java program that retrieves HTML pages from given URLs and automatically analyses and
classifies them according to DDC (see section 2). The DDC classmarks along with other accessible
metadata elements (see figure 5) are then represented in RDF using the Wolverhampton Core (wc)
schema (see Appendix A). The example pages have been selected from the top of a random range of
Yahoo categories as indicated. Note that the accession number is not set in the following examples
because the program is running as a stand alone application and not within the context of the search
engine.
Yahoo - Home : Social Science : Psychology
http://dir.yahoo.com/Social_Science/Psychology/Education

<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://www-nmcp.med.navy.mil/psychology/I1.htm">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>I1</wc:Title>
<wc:Abstract>Psychology Department Home Page Clinical Psychology
Internship Since 1990 the Psychology Department has offered a predoctoral
clinical psychology internship fully accredited by the American Psychological</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>psychology</rdf:li>
<rdf:li>psychological</rdf:li>
<rdf:li>association</rdf:li>
<rdf:li>adult</rdf:li>
<rdf:li>training</rdf:li>
<rdf:li>leadership</rdf:li>
<rdf:li>American</rdf:li>
<rdf:li>navy</rdf:li>
<rdf:li>naval</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>350 Public Administration and Military Science</rdf:li>
<rdf:li>158 Applied psychology</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>109</wc:Word_count>
<wc:Classification_date>11-Nov-98 14:53:32</wc:Classification_date>
<wc:Last_modified>07-Aug-98 14:55:04</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://spsp.clarion.edu/mm/RDE3/start/">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>Research Design Explained 3rd ed</wc:Title>
Page 13 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

<wc:Abstract>Aids for teaching research methods in psychology</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>computer</rdf:li>
<rdf:li>psychology</rdf:li>
<rdf:li>psychological</rdf:li>
<rdf:li>measure</rdf:li>
<rdf:li>experiment</rdf:li>
<rdf:li>experiments</rdf:li>
<rdf:li>research</rdf:li>
<rdf:li>learning</rdf:li>
<rdf:li>single</rdf:li>
<rdf:li>teaching</rdf:li>
<rdf:li>rights</rdf:li>
<rdf:li>writing</rdf:li>
<rdf:li>science</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>158 Applied psychology</rdf:li>
<rdf:li>150.724 Experimental research (Psychology)</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>205</wc:Word_count>
<wc:Classification_date>11-Nov-98 14:57:27</wc:Classification_date>
<wc:Last_modified>31-Aug-98 12:53:37</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
Yahoo - Home : Reference : Libraries : Library and Information Science : Institutes
http://dir.yahoo.com/Reference/Libraries/Library_and_information_Science/Institutes

<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://www.new-zealand.edu/infostudies/">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>Centre for Information Studies</wc:Title>
<wc:Abstract>Courses on Information Literacy Staff Administration
Current Courses Diplomas Certificates School based Courses Regional
Courses Holiday Courses Courses for non teaching staff Newsletter Web
Tutorial</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>librarianship</rdf:li>
<rdf:li>newsletter</rdf:li>
<rdf:li>education</rdf:li>
<rdf:li>school</rdf:li>
<rdf:li>administration</rdf:li>
<rdf:li>teacher</rdf:li>
<rdf:li>teaching</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>021 Relationships of libraries, archives, information centres</rdf:li>
<rdf:li>370 Education</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>34</wc:Word_count>
<wc:Classification_date>11-Nov-98 15:08:20</wc:Classification_date>
<wc:Last_modified>15-Sep-98 22:25:51</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
Page 14 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

<rdf:Description about ="http://www.mmu.ac.uk/h-ss/dic/">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>Department of Information and Communications MMU UK</wc:Title>
<wc:Abstract>The Department of Information and Communications at
the Manchester Metropolitan University UK Includes course research and
contact details</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>library</rdf:li>
<rdf:li>communications</rdf:li>
<rdf:li>school</rdf:li>
<rdf:li>university</rdf:li>
<rdf:li>science</rdf:li>
<rdf:li>management</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>380 Commerce, Communications, Transportation</rdf:li>
<rdf:li>027 General libraries, archives, information centres</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>7</wc:Word_count>
<wc:Classification_date>11-Nov-98 15:16:18</wc:Classification_date>
<wc:Last_modified>30-Jun-98 12:02:23</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
Yahoo - Home : Computers and Internet : Programming Languages
http://dir.yahoo.com/Computers_and_Internet/Programming_Languages/

<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://www.ampl.com/cm/cs/what/ampl/">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>AMPL Modeling Language for Mathematical Programming</wc:Title>
<wc:Abstract>FAQ BOOK SOLVERS PLATFORMS VENDORS CALENDAR MORE WHAT'S
NEW EXTENSIONS CHANGE LOG REPORTS NETLIB EXAMPLES CONTENTS HOME AMPL A
Modeling Language for Mathematical Programming Try</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>computer</rdf:li>
<rdf:li>programming</rdf:li>
<rdf:li>modeling</rdf:li>
<rdf:li>communication</rdf:li>
<rdf:li>mathematical</rdf:li>
<rdf:li>model</rdf:li>
<rdf:li>models</rdf:li>
<rdf:li>control</rdf:li>
<rdf:li>linear</rdf:li>
<rdf:li>nonlinear</rdf:li>
<rdf:li>discrete</rdf:li>
<rdf:li>interface</rdf:li>
<rdf:li>web</rdf:li>
<rdf:li>com</rdf:li>
<rdf:li>language</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>004.6 Interfacing and communications (Computer science)</rdf:li>
<rdf:li>005.1 Programming (Computer programming)</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>193</wc:Word_count>
<wc:Classification_date>11-Nov-98 15:22:31</wc:Classification_date>
<wc:Last_modified>08-Nov-98 23:39:21</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
Page 15 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://quimby.fla.com/Activities/Programming/APRIL/april.html">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>APRIL</wc:Title>
<wc:Abstract>Home page of the Network Agent Research Group within
Fujitsu Laboratories of America Inc</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>computing</rdf:li>
<rdf:li>programming</rdf:li>
<rdf:li>system</rdf:li>
<rdf:li>communication</rdf:li>
<rdf:li>model</rdf:li>
<rdf:li>internet</rdf:li>
<rdf:li>language</rdf:li>
<rdf:li>rights</rdf:li>
<rdf:li>interaction</rdf:li>
<rdf:li>america</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>003.5 Theory of communication and control (Computer Systems)</rdf:li>

<rdf:li>005.1 Programming (Computer programming)</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>46</wc:Word_count>
<wc:Classification_date>11-Nov-98 15:24:31</wc:Classification_date>
<wc:Last_modified>Not known</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
Yahoo - Home : Society and Culture : Religion and Spirituality
http://dir.yahoo.com/Society_and_Culture/Religion_and_Spirituality/

<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://www.theatlantic.com/election/connection/religion/religion.htm">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>Political Issues Religion</wc:Title>
<wc:Abstract>RELIGION Articles from The Atlantic Monthly 's archive
and related links Welcome to the Next Church by Charles Trueheart 1996
Seamless multimedia worship round the</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>intellectual</rdf:li>
<rdf:li>multimedia</rdf:li>
<rdf:li>vision</rdf:li>
<rdf:li>archive</rdf:li>
<rdf:li>cultural</rdf:li>
<rdf:li>school</rdf:li>
<rdf:li>copyright</rdf:li>
<rdf:li>service</rdf:li>
<rdf:li>religious</rdf:li>
<rdf:li>university</rdf:li>
<rdf:li>public</rdf:li>
<rdf:li>family</rdf:li>
<rdf:li>church</rdf:li>
<rdf:li>worship</rdf:li>
<rdf:li>god</rdf:li>
<rdf:li>christian</rdf:li>
<rdf:li>christianity</rdf:li>
<rdf:li>spiritual</rdf:li>
<rdf:li>religion</rdf:li>
<rdf:li>America</rdf:li>
<rdf:li>politics</rdf:li>
<rdf:li>political</rdf:li>
<rdf:li>rights</rdf:li>
<rdf:li>communities</rdf:li>
</rdf:Bag>
Page 16 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>027 General libraries, archives, information centres</rdf:li>
<rdf:li>210 Philosophy and Theory of Religion</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>243</wc:Word_count>
<wc:Classification_date>23-Nov-98 17:34:38</wc:Classification_date>
<wc:Last_modified>Not known</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
<?xml version="1.0"?>
<rdf:RDF
xmlns:rdf="http://www.w3.org/TR/WD-rdf-syntax#"
xmlns:wc="http://scit.wlv.ac.uk/~ex1253/wc/schema/">
<rdf:Description about ="http://www.inlink.com/~rife/religion.html">
<wc:Accession_no>0</wc:Accession_no>
<wc:Title>Dave's Controversial Religion Page</wc:Title>
<wc:Abstract>Dave's Controversial Religion Page The Monroe
Institute Spirit WWW site The Myth of the Historical Jesus Tibetan
Book of the Dead Faqir Chand The Unknowing</wc:Abstract>
<wc:Keyword>
<rdf:Bag>
<rdf:li>jesus</rdf:li>
<rdf:li>spirit</rdf:li>
<rdf:li>religion</rdf:li>
</rdf:Bag>
</wc:Keyword>
<wc:Classmark>
<rdf:Seq>
<rdf:li>210 Philosophy and Theory of Religion</rdf:li>
<rdf:li>290 Comparative Religion and Other Religions</rdf:li>
</rdf:Seq>
</wc:Classmark>
<wc:Word_count>76</wc:Word_count>
<wc:Classification_date>23-Nov-98 17:37:05</wc:Classification_date>
<wc:Last_modified>Not known</wc:Last_modified>
</rdf:Description>
</rdf:RDF>
Vitae
Charlotte Jenkins is a
Research Student at the
University of
Wolverhampton, UK. Her
research is concerned with
tools for information
Mike Jackson is Professor
of Data Engineering at the
University of
Wolverhampton, UK. He
is a Fellow of the British
Computer Society. He
Peter Burden graduated
with a BA in Mathematics
from the University of
Cambridge in 1964. He is
currently employed in the
School of Computing and
Jon Wallis works as an IT
consultant in the
pharmaceutical industry,
specialising in laboratory
automation and robotics.
Prior to this, Jon was a
Page 17 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

resource discovery on the
Web and in particular
automatic classification.
Charlotte graduated from
Oxford Brookes
University in 1995 with a
B.Sc. Joint Honours in
Computing and English
studies
obtained his first degree in
Computer Science at
Sheffield City Polytechnic
and his Masters at
Manchester University.
Mike has served on the
organising and programme
committees of numerous
database conferences
including BNCOD,
IDEAS, EDBT, ICDE, ER
and VLDB.
Information Technology at
the University of
Wolverhampton where he
teaches systems and
network programming and
is responsible for the
School's Unix based
systems. His research
interests include Internet
resource discovery and
cataloguing.
senior lecturer at the
University of
Wolverhampton. His
teaching was mainly in the
area of computer networks
and communication
systems, with research
interests in web search
engines and the
information management
issues of corporate web
sites.
Page 18 of 18Automatic RDF Metadata Generation for Resource Discovery
2003/05/08file://C:\Documents%20and%20Settings\hbrown\Desktop\8May\Automatic%20RDF%20...

This document was created with Win2PDF available at http://www.daneprairie.com.
The unregistered version of Win2PDF is for evaluation or non-commercial use only.