Making the Most of Controlled Vocabularies in Search Interfaces

1
Making the Most of Controlled Vocabularies in
Search Interfaces
presented byChris FarnumInformation Architect
ASIST IA Summit 2005 Montreal March 7, 2005
2
Outline

• Goal share techniques for making controlled
  vocabularies, indexing, and metadata accessible
  to searchers
• Intro/Background
• More like this
• Browsing Controlled Vocabularies
• Field Indexes
• Thesauri
• Search term suggestions

3
Introductions

• Chris Farnum
• Information Architect and Usability Design
  Product Manager, ProQuest
• previously worked with Argus Associates and
  Compuware
• 7 years IA experience
• LIS background U. Mich. SILS
• 5 years library public service
• crfarnum_at_yahoo.com
• chris.farnum_at_il.proquest.com

4
Background
ProQuest Search Redesign
After
5
Background

• Rich Metadata Available
• (ABI/INFORM Global and others)
• Subject
• Company
• Person
• Location
• Classification Codes
• NAICS/SIC
• Document Type
• and more

6
Background

• Guiding principles
• Find multiple ways to leverage metadata,
  especially facets.
• Place opportunities to use and learn about
  metadata in the path of the users workflow but
  dont require people to use them.
• Balance needs of expert and novice searchers.
• Make search tools context sensitive, so they
  appear when/where they are useful.

7
1 More like this

• Enable the berry picking strategy advocated by
  librarians.
• Let users see the metadata!
• Hyperlink terms for sideways searching.
• Allow users to combine terms and launch a search
  from the document level.

8
(No Transcript)
9
Ovid Wilson Business Abstracts
10
bibliographic metadata
faceted classification indexing
to combine terms
11
(No Transcript)
12
Best Practices

• More like this enables bottom up searching
• Multiple ways to implement more like this
• User chooses facets/terms
• Fully automatic

13
2 Browsing Controlled Vocabularies

• An alternative to searching.
• Not all users are search-dominant.
• Complements searching
• Browsing is better for some tasks

14
(No Transcript)
15
(No Transcript)
16
Browsing in ProQuest

• A separate area in the IA especially for
  browsing.
• Editorially managed taxonomy / directory
• Data-driven browsing powered by facetted index
  term associations (co-occuring terms)
• Browsing within special content collections
  (publications)

17
Editorially created taxonomy
18
Look up index terms
restrict to facets
narrow by co-occurring terms
19
breadcrumb navigation
to results
20
Channel
Category
Facetted drill-down is data-driven, but functions
as a taxonomy.
Location
21
Best Practices

• Offer novices assistance exploring and narrowing
  topics.
• Include a separate/alternate area for user who
  prefer browsing. Not everyone is
  search-dominant.
• Support data-driven facetted browsing when
  possible very flexible and always up to date.

22
3 Field Indexes

• Field indexes can be tools for helping users
  build queries.
• Enable users to find valid terms to use for
  searching specific fields.
• Assist users who dont know your search engines
  syntax.
• Can be data-driven or based on static lists.

23
OCLC FirstSearch
24
browseable indexes / pop-ups
25
abrasiv
26
(No Transcript)
27
Best Practices

• Make it simple - the look-up process and the
  steps to add terms to a search.
• Make tools optional but accessible.
• Types of tools
• Static / editorially created
• Data-driven A-Z lookups
• Tools for exploring facets / narrowing

28
4 Thesauri

• Got a thesaurus? Incorporate it into the search
  interface.
• Thesauri make can useful tools for building
  searches.
• Allow users to explore related terms and find
  approved terms.
• Experts will appreciate it, novice searchers can
  learn from it.

29
PubMed
30
(No Transcript)
31
(No Transcript)
32
Best Practices

• Primary user audience should determine how
  prominent the thesaurus and the complexity of its
  structure appears.
• Ways to incorporate
• Tool that assists in building a query.
• Separate search mode.
• Behind the scenes for example, the search
  engine automatically searches on synonyms

33
5 Search Term Suggestions

• Enable narrowing and browsing within the context
  of search.
• The majority of searches entered by users are 1
  or 2 keywords.
• No separate starting point required - users begin
  with the search box.
• Helps focus keyword queries by suggesting
  better/alternate terms and categories.
• Integrates searching and browsing.

34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
ProQuest Smart Search

• Database / context sensitive its aware of the
  database or combination of databases currently
  active
• Step 1 - Suggestions based on the users query
• topic terms (subjects, companies, people,
  locations) or pairs of terms AND-ed together
• Publications
• Step 2 Narrow as in topic guide with results
  visible.

40
1st results screen suggestions
41
Null results - suggestions
42
2nd screen narrowing using co-occurring index
terms
43
ProQuest Smart Search

• How it works
• Several methods used to analyze and match
  keywords with index terms.
• Exact matches
• Accounts for synonym relationships
• Co-occurrences with keywords and other index
  terms
• Associations with full text keywords in documents
  viewed/printed
• Editorially created associations
• Suggests paired combinations of terms if more
  than one keyword is entered.
• Suggests publications that are exact keyword
  matches and those with high numbers of matching
  index terms.

44
Best Practices

• Decide whether you are narrowing within the
  results set or making suggestions from across the
  while database.
• Integrate suggestions into the results screen.
• Include breadcrumbs.

45
Final Thoughts

• Use these models as inspiration, not a template.
• Choose and adapt them to your audiences needs.
• If you are investing in rich controlled
  vocabularies, these strategies can help to make
  the fruits of your labors more visible.

46
Live Demo (as time allows)

• A tour of ProQuests search UI and controlled
  vocabs.
• ProQuest (authentication required)
• Lets play stump Smart Search!
• Got a favorite search to try?
• For those playing at home, see the online promo
  for Smart Search
• http//www.proquest.com/division/pqnext/previews/S
  martSearch/

47
Special thanks to

• Information Architecture for the World Wide Web
  2nd Ed. (Lou Rosenfeld and Peter Morville)
• Search Engine Watch (Danny Sullivan) -
  http//searchenginewatch.com/
• Dan Hepp, John Law, and Hania Kutcher for their
  keys roles in designing the Smart Search
  technology the ProQuest user interface.

48
Contact

• Chris Farnum
• Information Architect
• ProQuest Information and Learning
• Ann Arbor, MI
• crfarnum_at_yahoo.com
• chris.farnum_at_il.proquest.com
• work phone [phone]
• url http//crfarnum.webhop.info