Topic Maps

1
Topic Maps

• Vijay Raghavan
• Distinguished Professor
• University of Louisiana at Lafayette -ULL
• The Center for Advanced Computer Studies -CACS

2
What are Topic Maps?

• An ontology technology
• based on a model of concepts and relations
• not hierarchical a network model
• starts from what the information is about
• Designed for information and knowledge
  organization
• based on the concepts from back-of-book indexes
• used to organize large sets of information
  resources
• Also used for information integration
• identity model used for automated merging
• information can be gathered from many different
  sources and integrated
• very powerful for portal integration

3
The 2-Layer Topic Map Model

• The core concepts of Topic Maps are based on
  those of the back-of-book index
• The same basic concepts have been extended and
  generalized for use with
• digital information
• Envisage a 2-layer data model consisting of
• a set of information resources (below), and a
  knowledge map (above)
• This is like the division of a book

4
(1) The Information Layer

• The lower layer contains the content
• usually digital, but need not be
• can be in any format or notation
• can be text, graphics, video, audio, etc.
• This is like the content of the book to which the
  back-of-book index refers

5
(2) The Knowledge Layer

• The upper layer consists of topics and
  associations
• Topics represent the subjects that the
  information is about
• Like the list of topics that forms a back-of-book
  index
• Associations represent relationships between
  those subjects
• Like see also relationships in a back-of-book
  index

6
(2) The Knowledge Layer Example
7
Linking the Layers Through Occurrences

• The two layers are linked
• together
• Occurrences are relationships with information
  resources that are pertinent to a given subject
• The links (or locators) are like page numbers in
  a back-of-book index

8
Types

• In topic maps topics can be typed, which provides
  considerable power for describing the world from
  which the topics are taken.
• This is a capability is missing from traditional
  classification techniques.
• Using this, one could create types and assign
  them to topics, and thus say that "topic maps"
  is-a "technology, "Norwegian" is-a "language",
  and "TMQL" is-a "query language", and so on.
• This is a very simple capability, but it is also
  very powerful.

9
Types

• Once explicit types have been provided it is
  possible to let the user perform searches such as
  "find 'paris', but show only 'places'", or to
  show lists of all cities, separate from other
  kinds of subjects.
• Without types there is no way to do this, since
  the necessary information will be missing.
• Since the types are themselves topics the creator
  of the topic map can choose which types to use
• As a result, the model is infinitely extensible
  and adaptable and can capture just about any kind
  of information.

10
Additional features Scope

• Scope can be attached to any name, occurrence, or
  association in a topic map. Basically, scope can
  be attached to anything you can say in a topic
  map. Scope allows you to qualify a statement, but
  still express it.
• Users can then choose to see all information in
  all scopes, or only those in particular scopes,
  basically tailoring their view of the world as
  they want to see it.

11
Additional features Scope

• Example
• If we have topic map about languages, and basing
  it on the ISO 639 and Ethnologue lists of
  language codes.
• We might want to record that ISO 639 assigns
  English to the Germanic language group, while
  Ethnologue considers it a West Germanic language.
• This can be done by scoping the association
  between English and Germanic with a topic
  representing ISO 639, and the association between
  English and West Germanic with a topic
  representing Ethnologue.
• Similarly, one might use scope to record that
  what Ethnologue calls Maldivian, ISO 639 calls
  Divehi.

12
Additional features URI

• URIs are used to identify subjects.
• A topic may have any number of subject
  identifiers (URIs) which identify the subject the
  topic is about.
• These URIs should point to resources which
  describe the subject to a human the resources
  are known as subject indicators.
• This allows subjects to be uniquely identified
  across topic maps and the entire web. For
  example, the URI http//www.topicmaps.org/xtm/1.0/
  core.xtmsuperclass-subclass uniquely identifies
  the subclassing association type.

13
Additional features Merge

• This unambiguous identification of subjects is
  used in topic maps to merge topics that, through
  these identifiers, are known to have the same
  subject.
• Two topics with the same subject are replaced by
  a new topic that has the union of the
  characteristics (names, occurrences, and
  associations) of the two originals. There is in
  fact a well-defined procedure for automatically
  merging topic maps based on this rule.
• The combination of globally unique identifiers
  and the merging procedure makes integration of
  diverse information sources and reuse of
  information very much easier.

14
OASIS

• OASIS is a published subjects activity, which is
  developing guidelines for how to create, publish,
  and maintain subject indicators intended for wide
  usage.
• One example of this is well-known URIs for all
  the countries in the world (based on the ISO 3166
  country codes), which will allow us to tell that
  when you say 'Norway' in one topic map and I say
  'Norge' in another, we mean the same thing. More
  on this in a future article.

15
Use case forskning.no

• Norwegian government portal to popular science
  and research information
• basically an online popular science journal
• owned by the Norwegian Research Council
• name means research.no
• Purpose
• present science and research information to young
  adults
• intended to raise interest and recruitment

16
The Dual Classification
17
Topic maps Standards

• Topic maps are an ISO standard, published as
  ISO/IEC 13250 in 2000.
• That standard defines the basic model and an
  SGML-based syntax for it, which uses HyTime for
  linking, and is therefore known as HyTM.
• TopicMaps.Org is an organization that was formed
  to create a more web-optimized topic map syntax
  based on XML and URIs.

18
Topic maps Standards

• TopicMaps.Org published its XML Topic Maps (XTM)
  1.0 specification in early 2001, and in October
  of the same year that syntax was accepted into
  the second edition of ISO 13250 as an annex.
• Today, XTM is the main topic map syntax and is
  supported by nearly all topic map tools.

19
How to create a topic map?

• For creating a topic map there are four main
  approaches
• Have humans author the topic maps manually. This
  usually gives very high-quality and rich topic
  maps, but at the cost of human labor. This is
  appropriate for some projects, while
  prohibitively expensive for others.
• Automatically generate the topic map from
  existing source data. This can give very good
  results if the existing data are well-structured
  .
• if not, there are various natural-language
  processing tools that might help.

20
Topic Maps Construction Approaches

• Automatically produce the topic map from
  structured source data like XML, RDBMSs, LDAP
  servers, and more specialized applications.
• To produce a topic map we need a text editor,
  and for automatic generation XSLT stylesheets can
  be used perfectly well. This won't be enough for
  all uses, of course, and therefore there is
  specialized software for topic map editing and
  automatic generation of topic maps.

21
Topic Maps Construction
22
Functionalities of the Construction Process

• A TM building approach must include the following
  functionalities
• Defining resources identifying resource types,
  adding, deleting, modifying, and merging
  resources.
• Identifying and maintaining concepts/topics.
• Identifying and maintaining relationships/associat
  ions between topics and relationship instances.

23
Functionalities of the Construction Process..

• Defining different views on a Topic Map including
  selected topics, relationships, and/or resources.
• Storing Topic Maps persistently either in
  standard XTM files or in databases.
• Merging Topic Maps.
• Importing/exporting Topic Maps.
• Including external resources
• Providing a user interface for search and
  navigation in the TM
• Evaluating and validating the resulting Topic
  Map.

24
TM construction approaches

• Topic Maps Building from Existing Data Sources
• TM construction from structured document content
  such as XML documents and web pages.
• from document metadata (RDF documents)
• From structured knowledge such as ontologies,
  existing database schema, learning repositories
• from unstructured documents.

25
TM construction from structured document content

• This approach intends to extract knowledge from
  web sites to help users find relevant information
  on the Web using clustering (unsupervised
  learning) techniques.
• The process starts by defining the profile of a
  TM (and later applying it to Web sites). Which
  characterize Topic Maps and help evaluate their
  relevance to users' information needs.
• Second, the analysis identifies topics that have
  no interest

26
from document metadata

• This aims to develop a framework and toolkit for
  auto-generating topic maps, called MapMaker, it
  consists of a set of configurable processing
  modules
• which are chained together according to the needs
  of each individual auto-generation application.
• The different modules have access to an RDF model
  that is constructed during processing.
• The RDF model is cleaned and extended, finally
  converted to a topic map.

27
From structured knowledge

• The architecture of the repository includes
  wrappers created to convert disperse knowledge
  structures into an integrated XML schema used in
  the repository
• the repository is implemented as a relational
  database (using MYSQL), an XML-enabled
  application server, a customized XML schema for
  Topic Maps, a set of XML stylesheets for
  transforming and displaying topic maps, a set of
  Java servlets and jsp programs to generate XML
  files dynamically from the database,

28
from unstructured documents

• The proposed approaches are based on different
  extraction techniques, namely learning techniques
  and Natural Language Processing techniques.
• This approach has focused on Inductive Natural
  Language Processing techniques to construct a
  Navigable Topic Map adapted to different users'
  viewpoints from free textual documents.
• In fact, by keeping track of words' association
  patterns, the system detect fluctuations in
  words' meanings which can reveal different points
  of view.

29
Putting Topic Maps in Context

• Topic maps are really an add-on to XML, something
  that adds extra value beyond what XML itself can
  do.
• Topic maps can be used without using XML at all.
• The two standards are similar without competing.
  They both have data models, interchange syntaxes,
  query languages, schema languages, and so on.
• Being developed for different purposes and doing
  different things. they can peacefully coexist and
  complement one another.

30
Putting Topic Maps in Context

• The relationship between RDF and topic maps is
  less obvious, however. Structurally, they are
  very similar, and their semantics are very close,
  although the distinctions in topic maps between
  base names, occurrences, and associations do not
  exist in RDF.
• At first glance it may appear that they are
  nearly the same, but on closer inspection it
  turns out that their respective communities think
  of the technologies in very different ways, and
  that features such as scope and merging actually
  make them rather different after all. Again, the
  conclusion seems to be that they are good for
  different things, and that there is room for
  both.

31
What should be used where?

• Generally, use XML for interchange and document
  contents, RDF for fine-grained metadata, and
  topic maps for making information findable and
  anything that is mostly about relationships

32
Benefits of topic maps

• A simple, intuitive model
• easy to teach to people, easy to apply
• focus on findability subjects and their
  relations
• Supports information architecture patterns
• taxonomies, thesauri, faceted classification
• synonym rings, best bets
• Formal structure
• supports advanced searching and querying
• can be exploited in many different ways, once
  created
• Easy to create web sites from
• the site structure flows from the Topic Maps
  ontology
• a simple natural model means you get an
  understandable site
• advanced search capabilities can be added easily

33
Summary

• Topic maps are not so much an extension of the
  traditional schemes as on a higher level. That
  is, thesauri extend taxonomies, by adding more
  built-in relationships and properties.
• Topic maps do not add to a fixed vocabulary, but
  provide a more flexible model with an open
  vocabulary.
• A consequence of this is that topic maps can
  actually represent taxonomies, thesauri, faceted
  classification, synonym rings, and authority
  files, simply by using the fixed vocabularies of
  these classifications as a topic map vocabulary.

34
Tools References

• The TAO of topic maps, the classic introduction
  to topic maps.
• TM4J is an open source topic map engine project
  in Java.
• Perl XTM is an open source topic map engine in
  Perl.
• tmproc is an open source topic map engine in
  Python.
• The Omnigator is a free (as in beer) topic map
  browser that can display any topic map. There's
  also an online demo.
• easytopicmaps.com is a wiki site about topic
  maps.

35
Tools References

• topicmap.com is a useful site about topic maps.
• XTM 1.0 is currently the most important
  specification. It has now been incorporated in
  the second edition of the topic map ISO standard.
  
• isotopicmaps.com tells you where the topic map
  standards are headed next.
• LTM, the Linear Topic Map notation, is a
  text-based syntax for topic maps that is easier
  to read and write for humans than XTM.
• Jan Algermissen maintains a registry of publicly
  available topic maps.