Topic Maps the GPS of the Web

1
Topic Maps the GPS of the Web
2
Outline

• Overview
• XTM
• Case Study 1
• Case Study 2
• Case Study 3
• Topic Map Software
• Topic Map Visualization

3
Overview

• Knowledge is becoming the key asset of
  organizations, but we find ourselves drowning in
  excesses of information. Given such a situation,
  how do we
• Find that needle of relevant information in the
  haystack of infosmog
• Capture and manage precious corporate memory
• Build a bridge between knowledge and the
  information
• Topic map can help with all three
• Revolutionize the ways in which we search for and
  navigate information
• Can model and represent knowledge in an
  interchangeable form
• Provide a unifying framework for representing
  knowledge and linking it with the information
  resources in which it is embodied

4
Overview (Cont.)

• Topic Maps (TM) is a technology to address the
  issue of semantically characterizing and
  categorizing documents and sections of documents
  on the Web w.r.t their content in other words,
  what topics or subject areas those documents
  actually address
• A TM act as a set of linked topics that index a
  document collection
• One can have multiple TMs indexing the same Web
  document collections
• TM can be viewed as information overlays on
  documents or arbitrary information resources
• TM acts as taxonomiesways of describing,
  classifying, and indexing an information space
  consisting of Web and non-Web objects

5
Overview (Cont.)

• Topic maps have a lot in common with the semantic
  networks used to represent knowledge in the field
  of artificial intelligence (topics and
  association).
• But they also add a new axis to the model of
  semantic networks -- that of occurrences -- that
  provides a bridge to the domain of information
  management.
• Knowledge management information management
• Knowing a thing versus simply having information
  about that thing

6
Overview (Cont.)

• The entities involved in an organization (people,
  roles, products, etc.) can be represented as
  topics (T)
• The complex and shifting relationships between
  those entities can be represented as associations
  (A)
• The documentation and other information resources
  that relate to them can be represented as
  occurrences (O).

7
An Opera Index
8
Features of Index

• Typographical conventions are used to distinguish
  between different types of topic (names of operas
  are shown in italic)
• Typographical conventions are used to distinguish
  between different types of occurrence (references
  to synopses are shown in bold)
• The use of see references handles synonyms by
  allowing multiple points of entry (by different
  names) to the same topic
• See also references point to associated topics
• Subentries provide an alternative mechanism for
  pointing out associations between different
  topics (e.g. between a composer and his works)

9
Features of Index (Cont.)

• A book may contain multiple indexes
• Homonyms can be distinguished through the use of
  explanatory labels following the names, e.g.
  Tosca (opera) and Tosca (character)
• The locators (page numbers) may contain modifiers
  that help distinguish between different types of
  occurrence, for example 54n for a footnote on
  page 54
• The nature of an occurrence (i.e., the way in
  which the information is pertinent to its
  subject) might also be shown using a subentry
  mechanism (clause, defined in, defined in
  glossary, used in production)

10
Topic Map Standards

• TM began in the pre-XML and pre-WWW era SGML
• TM today ISO 13250 is specified in terms of two
  different interchange syntaxes
• One based on an SGML DTD (that used the ISO19744
  HyperTime)
• XML TM syntax XTM
• 19 XTM elements

11
Components of the TM Standard

• SAM defines the formal data model of TM and its
  semantics in natural language
• Reference Model a mode abstract model of TM than
  SAM and to enable TM to semantically interoperate
  with other knowledge representation formalisms
  and Semantic Web ontology languages
• TMQL SQL-like language for query topic map
  information
• TMCL a database schema like capability to TM
  enabling constraints on the meaning to be defined
  for TM

12
Components of the TM Standard (Cont.)

• The products of the OASIS technical committees
  are intended to be layered onto the ISO13250
  standard's products
• The Published Subjects Technical Committee will
  define and manage published subjects, and
  establish usage requirements for these
• The XML vocabulary TC will define the vocabulary
  to enable TM to interact with existing and
  emerging XML standards and technologies
• The vocabulary will be defined as published
  subjects
• The GC TC will define geographical country,
  region, and language-based published subjects to
  ensure interoperability across geographical and
  linguistic boundaries

13
TM Concepts Topic

• Any distinct subject of interest for which
  assertions can be made (nearly everything in TM
  can become a topic)
• A topic is a representation of the subject
• A topic reifies a subject
• According to XTM, a topic acts as a resource that
  is a proxy (information representation) for the
  subject
• Each topic need a base name that reflects the
  intent
• ltbaseNamegt -- ltbaseNameStringgt
• What is a tomato???
• ltoccurrencegt -- ltresourceRefgt
• The base name string works fine for human, but
  maybe machines need some help ? need to say what
  the subject of my topic is more precisely

tomato1.xml
tomato2.xml
14
Topics
15
TM Concepts Occurrence

• A topic may be linked to one or more information
  resources that are deemed to be relevant to the
  topic in some way. Such resources are called
  occurrences of the topic
• Addressable (URI) ltresourceRefgt
• Not addressable and has a data value specified
  inline ltresourceDatagt

16
TM Concepts Subject

• Give the topic an identity that both machines and
  humans can understand
• Use a Published Subject Indicators (PSIs)
• ltsubjectIdentitygt -- ltsubjectIndicatorRefgt
• The way the subject of a topic is referred to is
  by having the topic point to a resource that
  express the subject
• The subject of the topic is represented by an
  occurrence of a resource, and it is the nature of
  that resource that determines the addressability
  of the subject
• resourceRef it constitutes the subject and is
  addressable
• subjectIndicatorRef it indicates the subject,
  not directly addressable

tomato3.xml
17
TM Concepts Subject (Cont.)

• A subject indicator is just a way of indicating
  subjects
• Topics are really the information representation
  of subjects
• A subject is indicated by defining a resource
• If two given topics use the same resource, then
  their subjects (identified or indicated by those
  resources) are identical
• XTM allows for a published subject indicator
  (PSI)
• A published subject is simply a subject that has
  general definition and usage and is identified by
  a specific published reference

18
TM Concepts Scope

• A ltbaseNamegt might be required to make sense to
  at least some humans
• Use ltscopegt to choose an appropriate ltbaseNamegt
• Default ltscopegt
• Depend on XTM application to choose
• lttopicRefgt -- point to an lttopicgt element that in
  turn has a subject
• Make the topic easier to read and write
• Make the topic map easier to maintain
• Occurrences can also be of different types,
  specified by the topicRef markup

tomato4.xml
19
Scopes
20
TM Concepts Association

• An association is the relationship between (one
  or more) topics
• ltassociationgt, ltmembergt, ltroleSpecgt
• An association is similar to the database notion
  of a relation, to the ontology notion of a
  predicate
• An association role specifies how a particular
  topic acts as a member of an association, its
  manner of playing in that association
• tomato5.xml, tomato6.xml

21
Associations
22
TM Concepts InstanceOf
tomato7.xml

• Topics can be categorized according to their kind
• Any given topic is an instance of zero or more
  topic types
• Topic types are themselves defined as topics by
  the standard
• Usage
• We can ask the topic map for all the dishes that
  have tomatoes as ingredients
• We can ask the topic map for all the desserts

23
TM Concepts variant (Cont.)

• Give a topic a variant name under a certain
  situation
• ltvariantgt, ltparametersgt, ltvariantNamegt,
  ltresourceDatagt
• ltresourceDatagt is a shortcut for ltresourceRefgt
• It would be foolish to have to create a file and
  a URI for every tiny piece of text in the whole
  topic map, so with ltresourceDatagt we allow text
  to be entered into the topic map document
  directly
• tomato8.xml

24
TM Concepts mergeMap

• The ltmergeMapgt element makes two or more topic
  maps to merge
• ltmenugt -- ltrecipegt -- ltpricegt
• Merge strategy
• All topics with the same name in the same scope
  are merged (a name-based merge)
• All topics with the same subject identity are
  merged (a subject-based merge)
• mergeMap enables the interchange of knowledge

25
Summary of TM

• Topic maps consist mainly of topics and
  associations
• A topic map is an overlay on information
  resources occurrence
• A topic is a stand-in, proxy, or surrogate for a
  subject PSI
• Topics have characteristics (names, occurrences,
  and roles played in association)
• The author controls the meaning of a topic map
  through topic characteristics and choices of
  subject
• Scopes in topic maps define the validity of
  associations and allow fine-tuning of merge
  operations

26
Summary of TM (Cont.)

• ltbaseNamegt
• ltbaseNameStringgt
• ltoccurrencegt
• ltresourceRefgt
• ltscopegt
• ltsubjectIdentitygt
• ltsubjectIndicatorRefgt
• lttopicgt
• lttopicRefgt
• ltassociationgt
• ltinstanceOfgt

• ltmembergt
• ltroleSpecgt
• ltmergeMapgt
• ltparametersgt
• ltresourceDatagt
• ltvariantgt
• ltvariantNamegt
• lttopicMapgt

27
TM Resources

• Entry points
• http//www.topicmaps.org/
• http//www.topicmaps.net/
• http//www.oasis-open.org/cover/topicMaps.html
• Sites for TM vendors and service providers
• http//www.infoloom.com/
• http//www.ontopia.net/
• http//www.semantext.com/
• http//www.empolis.com/
• http//www.cogx.com/
• http//globalwisdom.org/

28
Case Study 1Topic Maps in the Life Science

• Chapter 8 of XML Topic MapsCreating and Using
  Topic Maps for the Web

29
Overview

• Create a design for a new Web site that would
  allow learners all over the world to participate
  in the collection and representation of knowledge
  about the life science
• Develop a series of topic maps that will allow us
  to represent and navigate a large knowledge space
• If we are going to build a Web site where lots of
  different information can be captured using topic
  maps, it must begin with a knowledge structure
  that allow us to classify all related things

30
Linnaean Classification of Humans
31
The Five Kingdoms

• Kingdom ? Phylum ? Subphylum ? Class ? Subclass ?
  Infraclass ? Order ? Suborder ? Superfamily ?
  Family ? Genus ? Species

32
Some of the Phyla for the Animalia Kingdom
33
The Chordata Phylum
34
Creating Topic Maps for A Web Site

• User navigation
• Start with the big picture
• Drill down to more detail by selecting topic maps
  that are referenced as occurrences of some topic
  in the visible topic map
• Drill-down scheme
• A more detailed topic map is referenced as an
  occurrence of a particular topic in a less
  detailed topic map
• Developing the XTM Document
• Bottom-up
• Animalia TM ? FiveKingdoms TM

35
The Top-Level Topic Map
36
FiveKingdoms TM Pointing to the Animalia TM
37
Steps

• Create a shell for the Animalia topic map
• Create a shell for the FiveKingdoms topic map
• Create the TopicMap topic
• Since one of its occurrences will be an instance
  of a topic map, the bottom-up design approach
  suggests that we first define the TopicMap topic
  with a PSI
• Create the AnimaliaTopicMap topic

38
Steps (Count.)

• Create the Animalia topic
• Construct the topic that will server as a
  container for one or more occurrences of type
  TopicMap and for other associated information
• With this topic, we are now able to construct an
  occurrence that links the topic Animalia with the
  topic map Animalia
• Define an occurrence
• Select the Animalia topic and create a new
  occurrence
• Set the new occurrence as an instance of
  AnimaliaTopicMap
• Select the Topic Map Occurrence ? Animalia topic
  map

39
Creating the new Animalia Topic Map
40
Creating the new FiveKingdoms Topic Map
41
Creating the new TopicMap Topic
42
Creating the new AnimaliaTopicMap Topic
43
Creating the new Animalia Topic
44
Setting the InstanceOf Parameter in the New
Occurrence Editor Window
45
Selecting a Topic Map
46
Where are We Now?

• We create two topic maps
• Animalia
• FiveKingdoms
• Within the FiveKingdoms topic maps, we create
  three topics
• TopicMap
• AnimaliaTopicMap
• Animalia
• Create an occurrence, which offers the topic map
  Animalia as its resource reference

47
What's Next?

• Creating and maintaining Enterprise Web Sites
  with Topic Maps and XSLT
• Chapter 9 of XML Topic Maps Creating and Using
  Topic Maps for the Web
• The Cogitative Topic Map Web sites (CTW)
• Topic map source code (markup) that control Web
  site content and site maps
• XSLT stylesheets that control Web page layout and
  look-and-feel style
• The whole Web universe of resources referenced by
  XTM topic ltoccurrencegt resource locators

48
Case Study 2Linking Clinical Data Using XML
Topic Maps

• R. Schweiger, S. Hoelzer, D. Rudolf, J. Rieger,
  J. Dudeck
• Artificial Intelligence in Medicine 28
  (2003)105-115

49
Introduction

• Develop a search engine that allows indexing,
  searching and linking different kinds of clinical
  data
• Text matching methods fail to represent implicit
  relationships between data, e.g. HIV ??AIDS
• Topic maps provides a data model that allows
  representing arbitrary relationships between
  resources.
• Such relationships form the basis for a context
  sensitive search and accurate search results
• Relationships between the data are often hard
  wired in the application logic. XML, on the other
  hand, allows representing such relationships in a
  more flexible way.

50
Representing Relational Knowledge using TM

• Text matching relates search terms to resources
  and have limitations to identify relationships
  between the terms
• Search results are often inaccurate
• Need a data model that can represent arbitrary
  relationships between terms and other resources ?
  topic map
• AIDS phobia AIDS ??phobia
• AIDS phobia ??F45.2 (somatoform disorder)
• HIV ??AIDS

51
TM Relating "HIV" to "AIDS", and "AIDS" to
"image.gif"
TM enables a machine to reason that the term
HIV is indirectly related to the resource
52
"Human-immunodeficiency-virus" represents the
same concept as "HIV"
ltassociation id"F45.2"gt ltmembergt
lttopicRef xlinkhref"AIDS-phobia" /gt
lt/membergtltassociationgt
53
A few words about TM

• The most significant characteristic of a lttopicgt
  is the identifier, which specifies the topic as a
  fragment of the topic map and which allows to
  address the topic
• Topics have some implied meaning, also referred
  to as subject, that might be described somewhere
  else. The act of relating a topic to a subject is
  called reification
• In the long run, standards bodies will create
  so-called public subject indicators (PSIs), i.e.
  standard topics that are reused by topic map
  designers all over the world.
• Topic maps can also map between different topics
  representing the same subject ltsubjectIdentitygt

54
Context Sensitive Searching

• TM provides a flexible data model for
  representing arbitrary relationships between
  resources
• Need an inference method that use the given
  relationships
• Concept a meaningful relationship of terms
• Context sensitive searching ? concept macthing
• Association phase finds a set of concepts that
  relate the search terms meaningfully with each
  other
• Occurrence phase relates the resulting concepts
  to resources such as documents and images

55
Text Matching VS. Concept Matching
56
Context Sensitive Searching (Cont.)

• Concept matching allows a machine to understand
  precise queries and to produce accurate results.
• Search context A given set of terms
• Concept matching aims to find a context in a
  resource that relates the search terms
  meaningfully with each other, i.e. a resource
  context that matches the search context.

57
Semantic Linking

• TM can representclass-instance relationships
• As a result, we can categorize data and
  relationships between data
• Classified relationships, i.e. semantic links can
  be used to define customized search pathways

58
Semantic Network Linking ICD and DRG

• Enable a physician to enter diagnoses and codes
  and to find related information.
• Encode the semantic network using TM
• Search target "DRG"
• icdTitle?(synonym) ? icdTitle ? (has) ?
  icdCode ? (2drg) ? drgCode ? (drg) ?
  drgUri.

• ICD International Classification of diseases
• DRG diagnosis related groups

59
LuMrix Search for Diagnoses and Codes
60
Lurmix Search for Drug Information
61
Case Study 3Navigation and Interaction in
Medical Knowledge Spaces Using Topic Maps

• J. Beier, and T. Tesche
• International Congress Series 1230 (2001)pp.
  384-388

62
Introduction

• The medical occupation requires a widespread and
  up-to-date access to various information sources
• Text books, journals, guidelines, medical indexes
  (PubMed/Medline), selected internet sites, news
  groups, colleagues and medical experts
• An efficient use of these information resources
  in clinical routine is hampered by the
  heterogeneity and spatial distribution of the
  data, implying the necessity to utilize different
  retrieval techniques (libraries, telephone,
  internet).
• Using these conventional techniques, a
  comprehensive research requires a considerable
  amount of time, whichunder most circumstancesis
  not available

63
Introduction (Cont.)

• Limitation of medical search engines
• Internet search engines do not make use of a
  medical thesaurus but perform simple Boolean text
  pattern matching.
• Internet search engines collect each web page,
  disregarding its medical or non-medical content.
• The search string remains in the language it was
  entered. A translation to other languages of
  interest is usually not performed.
• In most retrieval tools, the context of the
  search (scope, aims) is not regarded and has to
  be expressed by the user specifying additional
  search terms.
• The vector space model most search engines use,
  calculates a hit ranking order based on word
  frequencies at document level. Unfortunately, an
  individual weighting of the specified search
  terms themselves is not possible.

64
Introduction (Cont.)

• Proposed method
• A knowledge-guided user front-end and an
  automatic generation of search engine queries
• The medical knowledge of MeSH (Medical Subject
  Headings) classification was transferred into a
  Topic Map
• The knowledge contained at each topic is utilized
  to control the search for documents or other
  information related to the query
• Enables an interactive navigation through topics
  of the medical (or another) domain.
• A graphical user-interface allows the fast and
  associative browsing in networks of themes.

65
Materials and Methods

• Topic Maps
• Topics (themes)
• Associations
• Occurrence TM connects nodes to related
  documents, images or other data
• For the developed IRS, special TM associations
  were chosen is-subclass-of, issuperclass-of,
  has-synonyms, has-preferred-term, is-related-to,
  is-definition, is-scope-note.
• MeSH classification (http//www.nlm.nih.gov/mesh)
• MeSH is the most used controlled vocabulary for
  document indexing
• http//www.nlm.nih.gov/databases/freemedl.html

66
Topic Map
Topic Maps combine a knowledge representation
with information resources
67
Proposed System

• Two-level architecture
• Knowledge organized by TM topics, associations
  and occurrences
• Documents described by their metadata and full
  text.
• User interface
• Navigation area for topic maps
• Description area of the selected topic
• Cards for document categories Guidelines,
  Lexica, EBM, Journals
• Hit list

68
Steps TM

• After entering a search string, the IRS
  determines a list of one or more topics that
  refer to this query concerning its title,
  synonyms and annotation in German/English
• According to the desired scope, the user chooses
  a topic and navigation within the topic map is
  started
• Ongoing from that starting point, the user
  interactively and graphically navigates through
  the network of themes, determining the topics
  that optimally fit to his or her demands.
• For the currently selected topic, additional
  information is displayed (MeSH code, definition
  and annotations, synonyms, translation).
• Using these topic commentaries, the system
  on-the-fly automatically generates a search
  query, expanding the search string with topic
  name, synonyms, translations, and definition.
• Pre-defined weighting factors at search string
  level control the impact of these different word
  groups for ranking the hits.

69
Steps Documents

• The document search space was divided into
  several categories (e.g. AHCPR guidelines,
  journals, selected internet sites).
• The categories were defined, extended and
  customized the systems administrator.
• The resulting hits are displayed in a manner
  known from internet search engines and are
  grouped according to their category

70
User Interface
71
Topic Map Software

• Chapter 10 of XML Topic MapsCreating and Using
  Topic Maps for the Web

72
Commercial Sources of TM Software

• Empolis http//k42.empolis.co.uk
• InfoLoom http//www.infoloom.com
• Mondeca http//www.mondeca.com
• Ontopia http//www.ontopia.net

73
Open Source TM Projects

• SemanText
• Construct, browse, and write rules and perform
  inference rules on topic maps
• Python based
• TM4J (http//tm4j.org/)
• XTM Programming
• A set of Java APIs for parsing, manipulating, and
  writing XTM

74
Open Source TM Projects (Cont.)

• Nexist (http//nexist.sourceforge.net)
• Java based XTM application
• Persistent Store HypersonicSQL (HSQL)
  (http//sourceforge.net/projects/hsqldb)
• GooseWorks (http//www.goose-works.org)
• Apache-licensed implementation of the graph-based
  data model for topic maps
• C/Python API
• TM for Javascript (TM4jscript)

75
Topic Map Visualization

• Chapter 11 of XML Topic MapsCreating and Using
  Topic Maps for the Web

76
Requirement for TM Visualization
77
Overview

• TM provide a bridge between the domains of
  knowledge representation and information
  management
• TM may be very large ? need a intuitive visual
  user interface to reduce users' cognitive load
• Different uses for Topic Maps
• If the user has a specific question ? query
  language (does not require visualization)
• Consider the relationships among objects ? more
  precisely
• If the user wants to simply explore a Web site, a
  TM can provide an overview so the user can decide
  where to start the exploration (require
  visualization)

78
Requirement for TM Visualization (Cont.)

• Two kinds of requirements for retrieving
  information
• Representation help users identify interesting
  sources
• Navigation help users access information rapidly
• Both representation and navigation are essential
  in a good visualization
• "the visual information-seeking mantra is
  overview first, zoom and filter, then details
  on-demand"

79
Representation Requirement

• Represent the whole topic map to help users
  understand it globally
• The overview should reflect the main properties
  of the structure
• Users should be able to focus on any part of the
  topic map and see all the dimensions they need
• Require the use of different levels of detail
  (generality/specificity)
• The position of topics on the visual display
  should reflect their semantic proximity
• Show all characteristics (topics, associations,
  scope)
• The representation should be updated in real time
  to enable user interaction

80
Navigation Requirement

• Navigation needs to be intuitive
• Free navigation should be kept for small
  structures or expert users
• Beginners prefer predefined navigation paths
• Expert users should be allowed to explore the
  structure freely

81
Visualization Techniques
82
Current TM Visualizations

• Most of them display lists or indexes from which
  users can select a topic and see related
  information
• Convenient when users' needs are clearly
  identified
• Usually the same as that on Web sites users
  click on a link to open a new topic or
  association
• Examples
• Ontopia Navigator (Omnigator)
• Empolis K42 application hyperbolic tree
• Mondeca's Topic Navigator graph representation
• UNIVIT 3D interactive TM visualization

83
Omnigator
84
Empolis K42 StarTreeView
85
3D Interactive Topic Map Visualization with UNIVIT
86
General Visualization Techniques Graphs and
Trees

• TM can be seen as a network of topics ? network,
  graph
• Graphs and trees are suitable for representing
  the global structure of topic maps
• Hyperbolic geometry allows the display of a very
  large number of nodes in a graph (efficient node
  positioning)
• Topics linked together by an association can be
  represented close to each other
• Topics of the same type or pointing to the same
  occurrences can be clustered
• Can represent the whole TM ? may become cluttered
  rapidly as the number of topics and associations
  increases

87
Example of a graph in 3D hyperbolic space
88
Graphs and Trees (Cont.)

• Different shapes and colors can be used to
  symbolize various dimensions of nodes and arcs
• The number of different shapes, colors, icons,
  and textures is limited
• Not suited for a TM containing millions of topics
  and associations

GraphVisualizer 3D
89
General Visualization Techniques Maps

• ET-Maps Internet home page categorization and
  searches
• Relative importance of each page according to the
  size of the corresponding zone
• May be used to represent topics and associations
• ThemeScape
• Topographical maps with mountains and valleys
• Documents with similar content are placed closer
  together
• Peaks appear where there is a concentration of
  documents about a similar topic (height of peaks)
• Valleys contain fewer documents and more unique
  content
• Topic labels reflect the major two or three
  related topics
• Different levels of detail

90
ET-Map
91
ThemeScape
92
General Visualization Techniques Virtual Words

• City metaphor
• A topic map a city
• Topics buildings (characteristics name, color,
  height)
• Association streets, bridges
• Topics and associations related to the same scope
  can belong to the same neighborhood
• Multiple dimensions of a topic map can be
  represented with this technique
• Navigate freely/guided tour walk/fly

93
Example of a Virtual City
94
Virtual City and A 2D Map
Occurrences and associated topics are displayed
in the bottom windows