Taxonomies for Information

1
Taxonomies for Information Knowledge Management
Architectures

• Denise A. D. Bedford, Ph.D.
• Senior Information Officer - Information
  Solutions Group Corporate Information Systems
• The World Bank Group
• Special Libraries Association DC Chapter
  February 4, 2003

2
Who Needs to Understand Taxonomies?

• Anyone who has been charged with the task of
  organizing information, regardless of the context
  Brick Mortar, KM system, ILS, Portal, Records
  Management System, .
• You may have been trained to work with a variety
  of Information Organization tools or trained in
  bibliographic control principles
• It is probably the case that these tools were not
  presented as taxonomies perhaps as rules, as
  record formats
• Professional literature also presents a
  confusing, often, simplistic view of taxonomies
• Assumes that a taxonomy is only hierarchical
• Fails to provide a clear distinction between the
  structures uses of classification schemes,
  subject headings, thesauri metadata

3
Information Management System Architectures

• The underlying architecture of a full
  bibliographic control system, records management
  system, a metadata repository or corporate
  information architecture is complex
• Taxonomies are essential structures in all
  information management systems
• Information managers, librarians, information
  architects, knowledge architects, records
  managers need to be able to
• Understand the different kinds of taxonomies
• Have sufficient familiarity with their purpose to
  select the right kind of taxonomy for an
  application

4
Taxonomy Basics

• There are four types of taxonomies
• Flat
• Hierarchical
• Faceted
• Network
• Some are explicit/visible, others are
  implicit/invisible
• There are significant design consideration when
  implementing each different type
• Lets review each quickly

5
Definition of a Taxonomy

• System for naming and organizing things into
  groups that share similar characteristics Jean
  Graef, Montague Institute

Taxonomy
Architectures
Applications
6
Taxonomy Basics

• Four types of taxonomies
• Flat
• Hierarchical
• Faceted
• Network

7
Flat Taxonomy Structure
Energy Environment Education
Economics Transport Trade
Labor Agriculture
8
Type 1 Flat Taxonomies

• Flat taxonomies group content into a controlled
  set of categories
• no inherent relationship among the categories in
  a flat taxonomy -- they are co-equal members of a
  single structure
• can move from one category to another without
  having to think about the relationship between
  them
• concept of a flat taxonomy may be counter
  intuitive to some
• Consider how often you use flat taxonomies
  everyday
• alphabetical listings of people in a directory of
  expertise
• a pull-down menu of country names or geographical
  regions
• simple alphabetical listings of product groupings

9
Designing Flat Taxonomies

• Flat taxonomies are easy to create
• Flat taxonomies do not require complex interface
  design and extensive usability testing
• We have learned from usability engineers how to
  implement flat taxonomies
• Flat taxonomies used for explicit information
  structures generally should consist of 30 or
  fewer categories
• More than 30 categories may be presented in a
  flat taxonomy, if the categories are intuitive to
  users (i.e. lists of countries, states,
  languages, etc.)

10
Explicit Flat Taxonomies

• Amazon.coms pull down list of product categories
  horizontal list of stores
• - http//www.Amazon.com
• Nordstom.coms alphabetical list of brand names -
  http//www.nordstrom.com
• Microsoft PowerPoints global functional menu
  pull down menu
• Water Resources Directory of Expertise list of
  keywords - http//www.nceas.ucsb.edu/exp/
• Bartleby.coms extensive picklists of reference,
  verse, fiction non-fiction listings --
  http//www.bartleby.com/
• CyberDeweys alphabetical index to sections -
  http//www.anthus.com/CyberDewey/Dewey_index.html

11
Implicit Flat Taxonomies

• Alphabetical list of water resource experts
• Content inventories listed alphabetically by
  author
• Rights management values (simple picklist)
• Information disclosure status values (simple
  picklist)
• Security classification scheme values (simple
  picklist)

12
Hierarchical Taxonomy
A hierarchical taxonomy is represented as a tree
data structure in a database application. The
tree data structure consists of nodes and
links. In an RDBMS environment, the
relationships become associations. In a
hierarchical taxonomy, a node can have only one
parent.
13
Type 2 Hierarchical Taxonomies

• Group content into two or more levels
• Resemble tree structures when they are fully
  elaborated
• Hierarchical categories typically have only one
  broader or parent category.
• Relationships among categories in hierarchical
  taxonomies have particular meaning
• Relationship between a top level category
  subcategory may mean group membership or
  refinement of the top category by a particular
  characteristic or feature
• Moving up the hierarchy means expanding or
  broadening the category
• Moving down the hierarchy means refining or
  qualifying the category

14
Explicit Hierarchical Taxonomies

• Yahoos Web Site Directory - organized as a
  subject hierarchy - http//www.yahoo.com
• Internet Public Librarys two-tier collection
  structure - http//www.ipl.org
• Librarians Index to the Internet mixed hierarchy
  of topics and resource types - http//www.lii.org/
  
• Ebays auction categories http//www.ebay.com
• CyberDeweys progressive disclosure of Dewey
  Decimal classes - http//www.anthus.com/CyberDewey
  /CyberDewey.html
• Albertsons Shop By Aisle grocery categories or
  Shop A to Z grocery product - http//www.albertson
  s.com/store/categores

15
Implicit Hierarchical Taxonomies

• Electronic news story published in XML NITF
  format
• International Press Telecommunications Council.
  News Industry Text Format. Version 3.1
  September, 2002. http//www.nitf.org/
• Classification schemes for topic areas
• Authority control lists for abbreviations full
  names (aliases)
• Records management hierarchical fileroom
  structures
• Cross source topic reference structures

16
Designing Hierarchical Taxonomies

• There is more than one way to implement a
  hierarchy
• Progressive disclosure of layers across sites or
  pages -- Ebay model
• Cascading or expanding menus -- United Nations
  web site
• Pop-up menus linked to stationary menus -- United
  Nations web site
• Category and subcategory labels in a multi-column
  display -- Nordstroms second level pages

17
Designing Hierarchical Taxonomies

• Hierarchical taxonomies should
• have content at every level -- empty categories
  present empty value to users
• be less than four levels deep in most cases
• be at least two categories for each branch in the
  taxonomy -- do not branch for a single category
• be sufficient content in each category to warrant
  existence
• balance breadth depth -- users must work harder
  to use a taxonomy three categories broad nine
  deep than to use one that is seven wide and two
  deep

18
Hierarchical Taxonomy Design Issues

• Hierarchical taxonomies should
• be balanced across each level of the taxonomy to
  provide users with a predictable experience
• be offset with search functions
• should never be displayed into flat structures
• be reviewed periodically

19
Facet Taxonomies
Faceted taxonomy represented as a star data
structure. Each node in the start structure
is liked to the center focus. Any node can be
linked to other nodes in other stars. Appears
simple, but becomes complex quickly.
20
Type 3 Faceted Taxonomies

• Resemble flat taxonomies when implemented, but
  have a different structure purpose than flat
  taxonomies
• There are no inherent relationships among
  categories in a faceted taxonomy like a flat
  taxonomy
• Resemble a star structure -- all facets pertain
  to the center object
• All categories in a faceted taxonomy relate to a
  single object -- may describe a property or a
  value, different views or aspects of a single
  topic

21
Type 3 Faceted Taxonomies

• An object may be
• electronic book -- each facet describes some
  aspect of the book - the author, the title, date
  of publication
• facets describe the countrys population,
  geography, economic system, political system,
  history,
• Each facet may relate to facets in other
  taxonomies -- a faceted taxonomy describing a
  book may also have a link to a faceted taxonomy
  that describes a country

22
Metadata as Faceted Taxonomy

• The primary implicit application of faceted
  taxonomies today historically is as implicit
  metadata records
• Traditionally, libraries have been the prime
  users or metadata
• Today, portals and e-business systems are primary
  metadata users
• Types of taxonomies that rely on metadata today
  
• IBMs product service catalog on the web
• User interest profiles
• Knowledge push or syndication profiles
• Selective dissemination of information or push
  profiles

23
Standards Based Metadata Schemes

• Dublin Core Metadata Element -
  http//dublincore.org/dcregistry/index.html
• GILS - Government Information Locator System -
  http//www.gils.net
• VERS - Victorian Electronic Records Strategy --
  http//www.prov.vic.gov.au/vers/welcome.htm
• MARC - Machine Readable Cataloging -
  http//www.loc.gov/marc/marcdocz.html
• UDDI - Universal Description, Discovery
  Integration of Business http//www.uddi.org
• TEI -- Text Encoding Initiative -
  http//www.tei-c.org
• ISAD(G) - International Standard Archival
  Description - www.ica.org/biblio/com/cds/isad_g_2e
  .pdf

24
Commercial Metadata Schemes

• Ebays auction item descriptions -
  http//www.ebay.com
• Amazon.coms product descriptions -
  http//www.Amazon.com
• Albertsons product descriptions -
  http//www.albertsons.com/store/
• Nordstroms product descriptions -
  http//www.nordstrom.com

25
Designing Faceted Taxonomies

• Most important design issue for faceted taxonomy
  is that it be suited to its purpose -- that it
  contain the facets that are needed, that their
  behavior is clear
• Characteristics of each facet should be defined
  fully and distinctly -- while all facets pertain
  to a common object, each has a distinct behavior
• Users should be able to manipulate facets
  distinctly -- it is important to define each
  facet exclusively, without overlap with other
  facets
• Most faceted taxonomies are implicit structures -
  when they are made explicit, they are generally
  presented as record or table formats

26
Network Taxonomies
A network taxonomy is a plex data structure.
Each node can have more than one parent. Any
item in a plex structure can be linked to any
other item. In plex structures, links can be
meaningful different.
27
Type 4 Network Taxonomies

• Organizes content into both hierarchical and
  associative categories
• May look like a computer network topology
• Many relationships among categories or nodes
• Relationships may have many different meanings
• Category may have more than one higher level
  category
• Any category in the taxonomy may be linked to any
  other category

28
Type 4 Network Taxonomies

• Examples of network taxonomies
• Topic maps or ontologies
• Thesauri -- !! A Thesaurus is NOT just a
  Hierarchy!!
• Semantic networks
• Implicit cross-walks for thesauri controlled
  vocabularies from different knowledge domains

29
Implicit Network Taxonomies

• Thesauri, concept maps semantic networks can be
  explicit or implicit
• Can be designed transparently into the knowledge
  management system as
• thesaurus facilitated search systems
• recommender engines (...if you liked this, you
  might also like this)
• vocabulary cross-walks from one source system to
  another
• topic map cross-walks from one knowledge domain
  to another

30
Explicit Network Taxonomies

• At one end of the scale we find simple, explicit
  network taxonomies such as topical taxonomies
  with See also references, or fully exposed
  thesauri
• At the other enc of the scale, we find more
  complex, explicit network taxonomies such as
  visual concept maps or visual semantic networks.
  Consider the following example of network
  taxonomies accessible on the web
• World Bank Groups Thesaurus http//www2.multite
  s.com/wb/
• UMLS Semantic Network - http//www.nlm.nih.gov/res
  earch/umls/META3.HTML
• Inxights Star Tree concept maps -
  http//eic.vestforsk.no/sitelense/eic.html

31
Designing Network Taxonomies

• Structure you use to maintain the network
  taxonomy may not work well for display -- may
  have to break it into one or more types of
  taxonomies to implement it
• Different kinds of relationships may be
  implemented in different ways - a single approach
  to display all types of relationships may not be
  effective
• Three-dimensional presentations are well suited
  to network taxonomy implemented, but should be
  usability tested with users
• Consider how the user will navigate a
  three-dimensional presentation
• Try to maintain a consistent level of granularity
  of categories - avoid mixing pre-coordinated
  subject headings or broad classes with concepts

32
Conclusions

• Every Information Professionals knowledge
  toolkit should contain a basic understanding of
  taxonomies their suitability to different
  applications
• Taxonomies are important building blocks in a
  full function information architecture
• Knowledge of taxonomies will be a skill in
  increasing demand as the need to organize
  information grows
• Need the organize information is growing
  exponentially, consistent with the amount of
  information produced, stored disseminated