Lecture 21: Facetted Classification

1
Lecture 21 Facetted Classification
SIMS 202 Information Organization and Retrieval

• Prof. Ray Larson Prof. Marc Davis
• UC Berkeley SIMS
• Tuesday and Thursday 1030 am - 1200 am
• Fall 2004

2
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

3
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

4
Controlled Vocabularies

• Vocabulary control is the attempt to provide a
  standardized and consistent set of terms (such as
  subject headings, names, classifications, etc.)
  with the intent of aiding the searcher in finding
  information
• That is, it is an attempt to provide a consistent
  set of descriptions for use in (or as) metadata

5
Hierarchical Classification

• Each category is successively broken down into
  smaller and smaller subdivisions
• No item occurs in more than one subdivision
• Each level divided out by a character of
  division (also known as a feature)
• Example
• Distinguish Literature based on
• Language
• Genre
• Time Period

Slide author Marti Hearst
6
Hierarchical Classification
Slide author Marti Hearst
7
Labeled Categories for Hierarchical Classification

• LITERATURE
• 100 English Literature
• 110 English Prose
• English Prose 16th Century
• English Prose 17th Century
• English Prose 18th Century
• ...
• 111 English Poetry
• 121 English Poetry 16th Century
• 122 English Poetry 17th Century
• ...
• 112 English Drama
• 130 English Drama 16th Century
• 200 French Literature

Slide author Marti Hearst
8
Faceted Categories

• Mutually exclusive
• Non-overlapping, distinct categories
• Relational
• Relations between facets, subfacets, and foci
  (elements) are not restricted to hierarchical
  generalization-specialization relations
• Composable
• Combined using grammars of order and relation to
  form compound descriptions

9
Faceted Classification Along With Labeled
Categories

• A Language
• a English
• b French
• c Spanish
• B Genre
• a Prose
• b Poetry
• c Drama
• C Period
• a 16th Century
• b 17th Century
• c 18th Century
• d 19th Century

• Aa English Literature
• AaBa English Prose
• AaBaCa English Prose 16th Century
• AbBbCd French Poetry 19th Century
• BbCd Drama 19th Century

Slide author Marti Hearst
10
Ranganathan

• PMEST Facets
• P(ersonality)
• WHO Types of things
• M(atter)
• WHAT Constituent materials
• E(nergy)
• HOW Action or activity terms
• S(pace)
• WHERE Where things occur
• T(ime)
• WHEN When things occur

11
Classical Facet Analysis

• Entity
• Kind
• Part
• Property
• Material
• Process
• Operation

• Patient
• Product
• By-Product
• Agent
• Space
• Time

12
Classical Facet Analysis

• What is being done?
• Entity
• Kind
• Product
• By-Product
• What are its parts?
• Part
• What are its properties?
• Property
• Material

• How is this achieved?
• Process
• By what means?
• Operation
• By whom?
• Agent
• Patient
• Where?
• Space
• When?
• Time

13
Classical Facet Analysis

• Nouns
• Entity
• Kind
• Part
• Patient
• Product
• By-Product
• Agent
• Adjectives
• Property
• Material

• Intransitive Verb
• Process
• Transitive Verb
• Operation
• Adverb
• Space
• Time

14
Semantic and Syntactic Relationships

• Semantic relationships
• Is-A (thing/kind, genus/species)
• Mammals
• Primates
• Humans
• Has-Parts
• Human
• Head
• Eyes

• Syntactic relationships
• Compounds
• Wheat harvesting wheat harvesting
• Object operation operation on object

15
Faceted Classification

• Clearly distinguishes between semantic
  relationships and syntactic relationships
• Semantic relationships
• Within a facet
• Containment relations
• Syntactic relationships
• Across facets
• Combinatoric relations
• Have a syntax for syntactic combination of
  semantic terms

16
Power of Facet Combinations

• The syntactic relations of faceted
  classifications enable a small controlled
  vocabulary to produce
• Many, many structured descriptions
• Complex, but formally structured descriptions
  using nested compound descriptions
• Descriptions for things we do not have words for

17
Example Objects
Red Plastic Glass
Blue Paper Straw
18
Project Team Facetted Classifications

• 007
• Personality
• Straw
• Glass
• Operation
• Drinking
• Slurping
• Sipping
• Material
• Plastic
• Paper
• Color
• Blue
• Red

• ARTery
• Color
• Size
• Material
• Weight
• Shape
• Radius/Circumference
• Density
• Volume/Capacity
• Function/Use
• Hardness/Softness
• Yin/Yang

19
Project Team Facetted Classifications

• Culture Feed
• Color
• Red
• Blue
• Material
• Plastic
• Paper
• Use
• Drink from
• Drink with
• Dimensions
• Circumference
• Height
• Diameter

• Picture Portal
• Color
• Red
• Blue
• Material
• Paper
• Plastic
• Use
• Containment
• Transport
• Shape
• Torus
• Planar
• Holes
• 0
• 1

20
Project Team Facetted Classifications

• F.U.N.
• Shape
• Color
• Material
• Rigidity
• Function
• Container
• Conduit
• Locale
• Weight
• Size

• MNM
• Functionality
• What it does
• What you can do with it
• Physical Properties
• Color
• Shape
• Material

21
Project Team Facetted Classifications

• pillBox
• Function
• Container
• Conduit
• Form
• Shape
• Cylinder
• Composition
• Paper
• Plastic
• Color
• Blue
• Red
• Size
• Tall and skinny
• Short and fat

• Team iTour
• Color
• Red
• Blue
• State
• Solid
• Non-porous
• Flexible
• Material
• Plastic
• Paper
• Geometry
• Cylindrical
• Hollow
• Function
• Container
• Drinking
• Sucking
• Blowing

22
Example Objects
Gray Metal Glass
Two Yellow Plastic Straws
23
Example Objects

• Function
• Form
• Shape
• Material
• Color
• Number

• Function Drinking
• Form
• Shape Cylinder
• Material Plastic
• Color Red
• Number 1

24
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

25
Faceted Classification Design

• Collect examples that need to be classified
• Identify candidates for facets and subfacets
• Test classification scheme on examples for facet
  orthogonality
• Order foci within facets
• Explicate grammar for ordering and combining
  facets and subfacets
• Test classification scheme on examples for
  combinatoric power
• Extend foci for comprehensiveness where
  applicable
• Create new facets and subfacets where needed
• Test classification scheme on new examples,
  especially boundary cases
• Iterate and refine throughout

26
Facet Guidelines

• Terms on the same level in the ontology should be
  of the same level and type
• Facets, subfacets, and foci should have a
  discernible order
• Use of capitalization and singular/plural forms
  should be uniform

• Sports
• Team Sports
• Baseball
• Football
• Basketball
• Solo Sports
• Marathon Running

• Sports
• Team Sports
• Baseball
• Football
• Basketball
• Solo Sports
• Marathon Running

27
Ordering Foci (Array)

• Simple to complex
• (Locomotions walk, run, jump, skip, hurdle,
  cartwheel)
• Common/popular to uncommon/unpopular
• (Vegetarian Pizza Toppings mushroom, onion,
  olive, artichoke, pineapple, pine nuts)
• Spatial, geographical, or geometric
• (Southwestern States California, Nevada,
  Arizona, New Mexico )
• Chronological, historical, or evolutionary
• (Dinosaur Eras Triassic, Jurassic, Cretaceous)
• Canonical (pre-established order)
• (Playground Counting Eenie, Meenie, Mynee, Mo)
• Alphabetical
• (Boys Names Al, Bob, Chuck, David, Ed, Frank,
  George, Harry)
• Size
• (T-Shirts Small, Medium, Large, XL, XXL)

28
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

29
Why Develop a Thesaurus?

• To provide a conceptual structure or space for
  a body of information
• To make it possible to adequately describe the
  topical content of information resources at an
  appropriate level of generality or specificity
• To provide enhanced search capabilities and to
  improve the effectiveness of searching (i.e., to
  retrieve most of the relevant material without
  too much irrelevant material)

30
Why Develop a Thesaurus?

• To provide vocabulary (or terminological) control
• When there are several possible terms designating
  a single concept, the thesaurus should lead the
  indexer or searcher to the appropriate concept,
  regardless of the terms they start with

31
Preliminary Considerations

• What is used now?
• Continue using an existing thesaurus?
• Ad hoc modification of existing thesaurus?
• Develop a new well-structured thesaurus?
• What is the scope and complexity of the subject
  field?
• What kind of retrieval objects or data will be
  dealt with?
• How exhaustive and specific is the desired
  description of objects?

32
Preliminary Considerations

• The scope and complexity of the field will
  provide some indication of the scope and
  complexity of the thesaurus
• It is better to plan for a larger and more
  comprehensive system than a smaller system that
  rapidly will become inadequate as the database
  grows
• Development of a good thesaurus requires a major
  intellectual effort as well as clerical
  operations like data entry and production of
  sorted lists

33
Development of a Thesaurus

• Term selection
• Merging and development of concept classes
• Definition of broad subject fields and subfields
• Development of classificatory structure
• Review, testing, application, revision

34
Flow of Work in Thesaurus Construction
35
1. Term Selection

• Select sources for the collection of terms
• Prearranged Sources
• Open-ended Sources
• Assign codes to each source
• Selection of terms
• For part of pre-arranged and for all open-ended
  sources
• Enter terms into database with all information

36
1.1 Kinds of Sources

• Prearranged Sources
• Existing descriptor lists, classification schemes
  thesauri
• This includes universal schemes like DDC or LCSH
• Nomenclatures of single disciplines
• Treatises on the terminology of a field
• Encyclopedias, lexica, dictionaries and
  glossaries
• Tables of contents of textbooks and handbooks
• Indexes of journals or abstracting journals
• Indexes of other publications in the field

37
1.1 Kinds of Sources

• Open-ended sources
• Lists of search requests or interest profiles
• Description of projects/activities to be served
  by the information retrieval system
• Discussion with specialists in the field
• Sample of documents in the field
• Ask users why and how these documents relate to
  the field
• Have documents indexed by experts in the field
• Lists of titles of documents in the field
• Abstracts and reviews of documents
• Your own knowledge

38
Selection of Sources

• Prearranged sources require less effort in
  gathering the material, and may already indicate
  some relationships between terms and concepts and
  relationships among terms
• Open-ended sources can reflect current
  terminology and may provide more complete
  coverage
• Choose a set of sources that are current, as
  complete as possible, and considered authoritative

39
Selection of Sources

• Each selected source is assigned an ID for
  tracking its use in the development of the
  thesaurus
• Useful when making decisions about which terms to
  prefer
• Useful for backtracking when questions arise
  (where did this come from?)

40
Selection of Terms

• Terms can be transferred directly from
  prearranged sources to the recording medium
  (cards or database)
• Have to decide which terms and references to
  include, or to take the whole source

41
Selection of Terms

• In open-ended sources you read through the source
  and pick out terms (i.e. words and phrases) that
  might be useful in retrieval or as references to
  other terms
• Alternatively, use keyword and phrase extraction
  software to create lists of terms and select from
  those
• Transfer selected terms to the recording medium
  (cards or database)

42
2. Merging and Development of Concept Classes

• Sort Term DB into alphabetical order
• First Round
• Merge information for identical terms, possibly
  pulling info from additional sources
• Second Round
• Merge synonyms or terms in the same concept class

43
3. Definition of Broad Subject Fields and
Subfields

• Define broad subject fields and sort terms into
  these broad fields
• Define subfields within each broad field and sort
  terms into these subfields
• Work out the detailed structure
• Select preferred terms
• Merge information for terms in the same concept
  class
• Repeat these steps
• For each subfield within a broad field
• And for each broad field
• Until all terms have been consolidated and
  preferred terms selected

44
4. Development of Classificatory Structure

• Produce preliminary version of classified index
  and update the working database
• Improve classificatory structure
• Reality check
• Produce and distribute a version of the
  classified index
• Distribute to users/experts

45
5. Final Stages

• Review
• Testing
• Application
• Revision

46
Review

• Discuss classified index with users/experts
• Select descriptors and checklist descriptors
• Assign notational symbols
• Produce main thesaurus and indexes

47
Review (cont.)

• Check cross references and insert where needed
• Produce test version
• Test by indexing
• Modify as needed
• Produce production version

48
Testing a Thesaurus

• Assign descriptors to a sample set of NEW
  documents (use enough to get an idea of any gaps
  in the thesaurus)
• Test retrieval using sample questions and seeing
  how effectively the thesaurus maps to the
  appropriate descriptor

49
Art and Architecture Thesaurus

• http//orange.sims.berkeley.edu/cgi-bin/flamenco/a
  a/Flamenco

50
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

51
Phone Project Assignments

• Photo Metadata Design (Assignment 6)
• Having your application and the overall project
  goals in mind, you will design a suitable
  metadata framework to use for annotating photos
  such that all photos would be accessible not only
  for the needs of your particular application, but
  also for the reusability of your photos and
  metadata by other applications.

52
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

53
Discussion Questions

• Paul Poling on Broughton
• What are the major inadequacies of 19th century
  classification systems which faceted
  classification overcomes?
• Some answers
• They don't "display very much in the way of
  internal logic, or fundamental structural
  principles ineffectual at addressing the
  specific problems of vocabulary they do not
  consider the precise relations between concepts
  multilingual switching difficult, particular in
  group/set names "fail to make adequate
  distinction between permanent hierarchical
  relationships, and relationships of syntactic
  association in complexes.   As a result,
  structures are not logical (since the analysis is
  not rigorous), positioning of compound subjects
  is not predictable (since no operating rules for
  combination are normally present), and retrieval
  is unreliable"

54
Discussion Questions

• Paul Poling on Broughton
• The author makes the somewhat startling claim
  that, "the fundamental thirteen categories have
  been found to be sufficient for the analysis of
  vocabulary in almost all areas of knowledge." 
  Are there any exceptions to this that come to
  mind?

55
Discussion Questions

• Paul Poling on Broughton
• Broughton later notes that some aspects of
  digital materials cannot be represented by the 13
  categories used for the BC2 system.  For use with
  our cameraphones, what are some categories that
  would need to be included?   More importantly,
  what is the minimum set of additional categories
  needed? 

56
Discussion Questions

• Paul Poling on Broughton
• Broughton states that, "There is no obvious way
  in which the core vocabulary can be dealt with by
  machines...the initial allocation of vocabulary
  tocategories must be carried out
  intellectually."  The author goes on to suggest
  that all but the initial category assignments can
  be done by a computer.  How feasible is the BC2
  system for the web, considering this requirement,
  when one considers the fairly rapidly expanding
  categories in so many fields of human
  knowledge?  

57
Discussion Questions

• Steve Chan on Broughton
• The category system used in BLISS/BC2 is based on
  a general ? specific ordering and on 13
  functional categories. How do you think that
  Lakoff's ideas of base level categories, and
  the importance of metaphor/embodiment relate to
  the categories chosen in Bliss/BC2?

58
Discussion Questions

• Steve Chan on Broughton
• Many of the relationships in the categories fall
  into types such as "is a kind of" or "is a part
  of". These are very similar to the predicates in
  WordNet. As a thought experiment, what would it
  take to interface WordNet into something like
  BC2, so that documents could be parsed for
  content and then automatically categorized? Would
  you want to let such a system generate the
  categories?

59
Discussion Questions

• Scott Fisher on Faceted Classification
• What are some different ways of ordering the
  facets within a classification notation?  When
  might one ordering be more appropriate than
  another?  Why might the result be especially
  important for non-electronic documents?

60
Discussion Questions

• Scott Fisher on Faceted Classification
• Why is it important that characteristics of
  division be mutually exclusive?  Explain what
  might happen if they are not.

61
Discussion Questions

• Morgan Ames on Vickery
• Though facets are a powerful tool for organizing
  information, they can be very time-consuming to
  define.  Vickery describes the creation of
  facets, starting with the analysis of terms used
  by a user group, then the sorting of the terms
  into facets, the development of facets (depending
  on how often they're used), the arrangement of
  the facets, and finally, the establishment of a
  notation for the facets.  Could one automate some
  or all of the process of defining facets for a
  particular area - say, an online community?  If
  so, which parts could be automated, and how?  If
  not, why not - what are the limitations of
  automation?

62
Discussion Questions

• Morgan Ames on Vickery
• How do the properties of facets compare with the
  properties of relational databases?

63
Discussion Questions

• Lilia Manguy on Thesaurus Construction
• The reading mentions thesauri being constructed
  for institutions. What are some examples of
  institutions with specialized thesauri? Why were
  they deemed necessary?

64
Discussion Questions

• Lilia Manguy on Thesaurus Construction
• In our field, what are some scenarios in which a
  thesaurus would need to be constructed? How would
  you determine who would be your expert
  consultants? Who would you choose?

65
Discussion Questions

• Lilia Manguy on Thesaurus Construction
• Using the process outlined in the reading for
  constructing a thesaurus, how would you qualify
  whether your thesaurus is good or bad?

66
Discussion Questions

• Christine Jones on Card Sorting
• Considering the "vocabulary problem" laid forth
  in "The Vocabulary Problem in Human System
  Communication," by Furnas et. al., do you think
  the card sorting technique is an effective
  approach for categorizing information for the
  SunWeb Intranet, i.e. do you think menus and the
  search function contain vocabulary users will
  understand? Would yourecommend any other tools
  for the user to increase their understanding of
  the SunWeb information space?

67
Discussion Questions

• Christine Jones on Card Sorting
• Usability studies including card sorting, icon
  intuitiveness testing, card distribution to
  icons, and thinking aloud walkthrough were
  performed and the results were based in part on
  subjective interpretation. For example, instead
  of depending on formal statistics, eyeballing the
  data was used and when deciding whether to keep
  icons, the user interface designers made the
  final decisions. Do you think this level of
  subjective interpretation was justified for a
  project of this nature? What (if any) changes
  would you make to this approach if the project
  was a redesign or design of Sun's external
  Website?

68
Discussion Questions

• Carrie Burgener on Flamenco
• How do the search and browse functions used by
  Flamenco compare to Bates Berry Picking Model?

69
Discussion Questions

• Carrie Burgener on Flamenco
• The examples in the article were collections of
  images that had existing metadata associated. It
  has been presented in IS203 that people take
  pictures and generally do not organize them. How
  can the UI design of Flamenco be applied to photo
  annotation?

70
Discussion Questions

• Carrie Burgener References for Flamenco
• PhotoCompas tool using Flamenco interface
• http//shark.stanford.edu4230/cgi-bin/flamenco/mo
  r_full/Flamenco?usernamedefault
• Presentation by Professor Hearst
• http//bailando.sims.berkeley.edu/talks/dli02.ppt
• Different article
• http//www.sims.berkeley.edu/hearst/papers/cacm02
  .pdf

71
Agenda

• Facetted Classification
• Traditional vs. Facetted Classification
• Designing Facetted Classifications
• Thesaurus Design
• Assignment 6
• Discussion Questions
• Action Items for Next Time

72
Homework (!)

• Assignment 6
• Due Thursday, November 18
• Read
• Textbook Organization of Information Chapters
  3-5 (Taylor)
• Chitra 3
• Shufei 4
• Jaime 5