Classification and Taxonomy

1
Classification and Taxonomy
Greg Argo
2
Brief origins of the organization of information

• Large amounts of information became difficult to
  store and retrieve.
• Although the classes used vary wildly across
  cultures, grouping based on the class level is
  nearly universal.
• Organizational structures provide the context in
  which humans transform information into
  knowledge.
• Its not just handy, its essential.

3

• Humans classify with a pronouncedly mental
  scalpel that helps us carve discrete mental
  slices out of reality because reality is not
  made up of insular chunks unambiguously separated
  from one another by sharp divides, but, rather,
  of vague, blurred-edge essences that often spill
  over into one another.
• -Eviatar Zerubavel (1991)from The fine line
  Making distinctions in everyday life

4

• Cognitive scientists have noticed that much of
  our mental commerce with an environment deals
  with classes of things rather than with unique
  events and objects.
• -Mark Stefik (1995) from Introduction to
  knowledge systems
• For example, the people seen below could probably
  all be placed in both the class Cognitive
  Scientists and the class Nerds. Can you think
  of other possible classes? Possible
  relationships? Clinical vs. academic cognitive
  scientists? Beards and nerds?

5
Why consider classification and taxonomy together?

• Both are methods for grouping objects or ideas
  sharing useful, although sometimes superficial,
  similarities
• Both group to make retrieval easier
• Both are very basic and pervasive elements of
  information architecture
• It is often difficult to tell them apart
• It is often unnecessary to tell them apart

6
Why tell them apart then?

• To become knowledgeable about the different
  limitations and possibilities in their
  interaction
• Differential demand on and payoff for users
• It is important to understand the specific
  qualities by which each can achieve
  organizational objectives

7
Specific qualities presented as keywords and
key-dichotomies

• Organization
• Retrieval
• Controlled vocabulary/thesauri
• Ambiguous vs. Exact
• Searching vs. Browsing

• Content-based vs. User-based
• Descriptive vs. Navigational
• Precision vs. Recall
• Structures vs. Applications
• Concise vs. Broad

8
Classifications, Taxonomies, and Ontologies -
Classifications

• Relationships expressed are not essential, but
  are based on arbitrary, external attributes
  (color, genre, format, geography, subject,
  alphabetical order)
• Created broadly from the top-down, based on
  conceptual frameworks
• Created by subject experts
• Usually dont change significantly after their
  creation
• Generally applicable to specific domains

9
Classifications, Taxonomies, and Ontologies -
Taxonomies

• Relationships expressed are usually essential,
  based on internal properties of the related
  pieces of information
• Created concisely from the bottom-up from actual
  content
• Created by multidisciplinary teams
• Are process-oriented, and so are updated
  frequently
• Oftentimes can be used and reused in different
  situations and environments
• Relationships commonly represented hierarchically
• Can be include many classifications connected
  together

10
Example of internal properties of taxonomic
relationship

• All zippers are clothes fasteners
• Not all clothes fasteners are
• zippers
• Because of the essential nature
• of their relationship, zippers is
• a sub-class of clothes fasteners,
• and clothes fasteners is a
• superordinate class of zippers

11
Taxonomic Hieracrhy

12
Classifications, Taxonomies, and Ontologies -
Ontologies

• Like taxonomies, relationships expressed are also
  essential
• Scope is more overarching due to inclusion of
  supplemental information
• Descriptions and definitions of concepts and
  their corresponding relationships
• Can include many sub-class taxonomies connected
  together

13
Classifications, Taxonomies, and Ontologies

• Classifications guide users to a body of
  information
• Taxonomies guide users through a body of
  information
• Ontologies guide users in becoming proficient in
  the retrieval of and understanding of a
  particular body of information

14
Classification

• To classify something is to identify it as a
  member of a known class
• On the Web, information architects organize
  classification schemes into either exact or
  ambiguous schemes
• Classification problems begin with data and
  identify predetermined classes as solutions

15
Exact classification schemes

• Items are categorized mutually exclusively
• Useful to users who know exactly what they are
  looking for
• By definition, are easier to create and maintain
  than ambiguous schemes
• Alphabetical, chronological, geographical

16
Alphabetical schemes

• Directories and lists
• User must have a good idea of what they are
  searching for and be able to spell it
• On the Web, usually utilized deeper in the scheme
  inside of sub-sites

17
Chronological schemes

• Have an intuitive advantage for users because
  they are organized in the same linear scheme in
  which humans experience the dimension of time
• Yearbooks, historical sites, and news headline
  sites
• Ebay offers results organized by a few different
  types of chronologies

18
Geographical schemes

• Have intuitive appeal to rich spatial faculties
  and needs of users in their experience of reality
• Geographical divisions coincide with governing
  bodies which restrict and encourage behaviors
  through law and language
• Requires knowledge of geographical divisions and
  map reading on the part of the user

19
Ambiguous classification schemes

• Items are categorized into intellectually
  meaningful groups
• Useful to users who dont know quite what
  information they are searching for
• Facilitate iterative, serendipitous learning
• Audience-based, Subject-based, Task-based
• Each should be based on scheme specific research
  and development processes (e.g. user and task
  analyses)

20
Audience-based classification schemes

• Makes sense if the informational domain caters to
  clearly delineated audiences
• Homepage becomes a filter that leads to sub-sites
  organized some other scheme
• Suggests customization/personalization
• Recommendations are sometimes powerful, sometimes
  failures

21
IA research for audience-based classification
schemes

• Map services and applications to their
  appropriate group
• Discern what types of technology-use are
  associated with specific populations
• Find points of overlap between audience
  categories
• User research sessions, usage statistics, search
  log analysis, focus groups, critical incident
  reports

22
Subject-based classification schemes

• Most immediately recognized are the library
  classification schemes (DDC, LC)
• When used in IA, they generally work best when
  hybridized with other types of schemes
• Are challenging to implement because different
  words, symbols, and idioms mean different things
  to different people
• Breadth of subjects included should be decided
  early on because these parameters will affect
  much of the rest of the IA and content work for
  the Web site

23
IA research for subject-based classification
schemes

• Solicit development team to write down each
  content item that will be part of site
• IAs perform card sorting exercise to establish
  initial subject categories
• Take it to the user
• Further card sorting
• Survey with questions about navigation
• Continually refine

24
Task-based classification schemes

• Useful for action and transaction related Web
  sites
• Rarely drive a Web site on their own, but are
  typically embedded deeper as part of a hybrid
  scheme
• Desire of businesses to remove labor costs will
  likely increase their ubiquity

25
IA research for task-based classification schemes

• The field of usability arose from the need to
  research the success and value of tools and their
  applications
• Traditional usability tests are a good fit
• Analyses of video-taped sessions, navigation
  logs, heuristic reviews, surveys, critical
  incident reports

26
Taxonomies

• Information architects have two major types to
  utilize descriptive and navigational
• They contrast well and each excels for different
  organizational and user needs
• Central ideas include creating hierarchies,
  controlled vocabularies, and variant/preferred
  term and synonym relationships
• Build on classifications by supporting
  applications and many different types of content,
  including images, email, search engines, process
  funnels, and site registration

27
Descriptive taxonomies

• Operate outside of a users immediate awareness
• Supplement information retrieval during keyword
  searching
• IAs create controlled vocabularies and synonym
  rings which they use to maintain consistency
  across applications and departments
• By analyzing emerging content and search logs,
  IAs maintain currency and map alternative
  terminology used by searchers back to the
  preferred form

28
Controlled vocabularies in descriptive taxonomies

• Done by attaching tags to content with metadata
  derived from controlled vocabulary usage logs
• The resulting thesaurus with related and variant
  terms makes a descriptive taxonomy more robust

29
Using the controlled vocabulary to increase
recall or precision

• A users search can be expanded to increase
  recall by mapping the search term to its variants
• Or a users search can be narrowed to increase
  precision by mapping a users term to the
  preferred term in the controlled vocabulary

30
More about descriptive taxonomies

• Created from the bottom-up
• Are called descriptive because they are derived
  directly from the content that is being used
• Data management vocabularies allow workers in
  disparate domains to report information using the
  same terminology
• Makes it easier for management to mine
  information from this data in the future

31
Navigational taxonomies

• Have a lot of overlap with exact and ambiguous
  classification schemes
• In contrast to descriptive taxonomies,
  navigational taxonomies command the users
  conscious awareness
• Allow the user to guide the seeking process
  themselves by browsing instead of searching

32
Navigational taxonomies contd

• Created from the top-down based on mental models
  of users
• Hierarchical structures visually imply sequences
  of events and relationships
• These relationships provide context similar to
  words in a sentence
• Works best when users are unsure of what they are
  seeking

33
Breadth vs. Depth

• Breadth is how many categories are contained in
  each level
• Depth refers to how many levels are contained in
  the hierarchy
• Too broad and shallow causes user too many
  choices and not enough content
• Too narrow and deep causes user to click more
  than they will stand for
• It is best to err on the side of broad and
  shallow to allow for add-ons and to avoid
  restructuring the home page

34
Summary

• Distinction is more pronounced in theory than in
  practice because both are essentially controlled
  vocabularies structured by logical relationships
• Generally, as one moves from classifications to
  taxonomies to ontologies, the structures,
  relationships, and supplemental descriptions
  become more complex

35
Summary contd

• Since humans seem to perform all three of these
  innately, it matters less what they are called
  than how their elements can be tailored to
  specific scenarios to improve retrieval of
  information, consistency of communication, and
  creation of knowledge

36
References

• Adams, K. (2000). Immersed in structure the
  meaning and function of taxonomies.Internetworking
  , 3.2. Retrieved October 25, 2004
  from http//www.internettg.org/newsletter/aug00/a
  rticle_structure.html
• Brown, J., Duguid, P. (2002). The social life
  of information. Boston Harvard Business School
  Press.
• Conway, S., Sligar, C. (2002). Unlocking
  knowledge assets. Redmond, Washington Microsoft
  Press.
• Edols, L. (2001).Taxonomies are what? FreePint,
  97, 9-11. Retrieved October 25, 2004 from the
  FreePint Web site http//www.freepint.com/issues/
  041001.pdf
• Goodall, G. (2003). Business taxonomies and
  bibliographic objective Facetation. Retrieved
  October 25, 2004 from http//www.deregulo.com/fac
  etation/pdfs/businessTaxomies_goodall.pdf
• Nielsen, J. (2001). Designing web usability.
  Indianapolis, IA New Riders Publishing.
• Rosenfeld, L., Morville, P. (2002). Information
  architecture for the World Wide Web. Cambridge
  Sebastopol, CA O'Reilly.

37
References contd

• Shank, P. (2004). Get organized or get lost.
  OnlineLearningMag. Retrieved October 25, 2004
  from http//www.onlinelearningmag.com/onlinelearn
  ing/magazine/article_display.jsp?vnu_content_id11
  08349
• Stefik, M. (1995). Introduction to knowledge
  systems. San Francisco Morgan Kaufmann.
• Svenonius, E. (2001). The intellectual foundation
  of information organization. Cambridge, MA The
  MIT Press.
• Taylor, Arlene G. (1999). The organization of
  information. Englewood, CO Libraries Unlimited.
• van Duyne, D. K., Landay, J. A., Hong, J. I.
  (2003). The design of sites. Cambridge
  Addison-Wesley.
• van Rees, R. (2003). Clarity in the usage of the
  terms ontology, taxonomy and classification.
  CIB73 2003 Conference Paper. Retrieved October
  25, 2004 from http//vanrees.org/research/papers/c
  ib2003.pdf
• Zerubavel, E. (1991). The fine line Making
  distinctions in everyday life. New York
  Free Press.