You want me to do what

1
You want me to do what?

• The Transition from Distributed to Unified,
  Federated CombinedSearching
• A somewhat light-hearted view offinding a new
  model for accessto scholarly resources
• Senior Staff
• May 10, 2005

2
Weve come a long way

• We have come a long way since 1989 when we
  introduced our first system for searching
  indexes.
• Today we offer a vast array of electronic
  resources using many systems.
• The expectations of our users have changed as
  well.

3
Abstract Index Searching

• Today we offer students and faculty
• 200,000,000 citations
• 500 titles
• 350 are from 50 systems with multiple databases
• 150 are direct from the vendor
• Over 200 interfaces
• All of them are different!

4
You want me to do what?

• Do you expect me to do all that search the
  library stuff? Like, get real! Id rather use
  Google.
• If you were a student today,wouldnt you have
  the samereactions to what we expect?

Fear
Frustration
5
Distributed Searching
Many interfaces
Many search engines
Many databases
Many search results
6
Unified Searching The Impossible Dream

• Our goal is to provide one place for students and
  faculty to go for all scholarly resources.For
  many reasons this is The Impossible Dreambut
  we are working toward this goal

7
Unified Searching
Single interface
Single search engine
Many databases
Single search result
8
Unified Searching

• is helping turn these reactions into these

9
Unified Searching

• We will never achieve fully unified searching
• No aggregator has everything
• Many database vendors wont allow local loading
• Scholars Portal Search is based on this model
• but it will never have everything
• at best it will have only
• 25 of our titles
• 50 of our citations
• What can we do about the rest?

10
Federated Searching

• Bringing databases into a uniform framework
• Searching everything with a single interface
• Federated searching comes close to doing this
• What is federated searching?
• A system used to
• simultaneously execute searches across many
  different information sources
• collect and collate the responses
• present the results in a single list

11
Federated Searching

• Can help turn the grumpiest of users from this
  

• to this

12
Federated Searching
Search native interface
Single interface Search agent
Many search engines
Many databases
Single search result
13
Unified vs Federated Searching
Unified
Federated
Single database with a unified indexing structure
Many different databases each with its own
indexing
Database Indexing
There is a single result set for the database
There is a merged result set with the top ?
citations from each database
Search Results
All citations are ranked using the same
algorithm
Ranking is done within each database with the
ranking algorithm of the system being
searched Undisplayed results from one database
may be more relevant to the query than the
displayed results from another database
Relevancy Ranking



Deduplication of Records

Search Performance
14
Unified vs Federated Searching
Database Relevance A Highly B Moderately C
Slightly
Unified
Federated
Single database with a unified indexing structure
Many different databases each with its own
indexing
Database Indexing
There is a single result set for the database
There is a merged result set with the top ?
citations from each database
Search Results
All citations are ranked using the same
algorithm
Ranking is done within each database with the
ranking algorithm of the system being
searched Undisplayed results from one database
may be more relevant to the query than the
displayed results from another database
Relevancy Ranking


Deduplication of Records

Search Performance
15
Unified vs Federated Searching
Unified
Federated
Single database with a unified indexing structure
Many different databases each with its own
indexing
Database Indexing
There is a single result set for the database
There is a merged result set with the top ?
citations from each database
Search Results
All citations are ranked using the same
algorithm
Ranking is done within each database with the
ranking algorithm of the system being
searched Undisplayed results from one database
may be more relevant to the query than the
displayed results from another database
Relevancy Ranking

Because data is loaded in a single data
structure with the same database definition, the
masking of duplicate records is reliable
Because data is loaded from different sources
with different database definitions, the masking
of dupli-cate records less reliable
Deduplication of Records
Searching is done on a single system on which
performance can be managed
Searching is done on multiple systems on which
performance is managed by each target system
Search Performance
16
How Federated Searching Works
Target Database
User enters query
Target system searches database
Search agent sends query to target
Target returns top citations
Search agent collects results
Search agent sends results to user
Search agent merges results
Search agent dedups results
Search agent ranks or orders results
17
Federated Search Products

• Product Vendor
  Based on
• AGent Auto-Graphics
• Chameleon VTLS Muse Global
• ERA Endeavor Muse Global
• Horizon Portal Dynix WebFeat
• MetaLib Ex Libris
• MultiSearch CSA Muse Global
• SingleSearch Sirsi Muse Global
• WebFeat 3 WebFeat
• ZPORTAL Fretwell-Downing

18
Muse Global

• Muse Metasearch
• provides a single action for simultaneous
  searching across a diverse set of content sources
• gives searchers a single sign-on for
  authenticated access to local and remote
  resources, hiding the complexity of searching
  multiple repositories
• returns results in a consistent, uniform format
  which can then be sorted, filtered and
  de-duplicated
• The Muse Global market is not libraries it is
  information system vendors

19
CSA MultiSearch

• We have licensed CSA MultiSearch for use by the
  University of Toronto Libraries
• Brings non-Scholars Portal resources into the
  framework of Scholars Portal Search
• Being considered by other Ontario universities
• Will probably become a Scholars Portal initiative

20
CSA MultiSearch

• Provides federated access to resources
• Is powered by the CSA Illumina platform

21
Targets Connectors

• Muse Global writes the connectors that link the
  Illumina interface and the target information
  sources
• There are connectors for more than a thousand
  targets
• Targets may be
• Web servers
• Z39.50 servers
• Other information sources

22
UofT Targets
We must now select our MultiSearch targets

• Identified targets
• Scopus
• Sirsi
• Library web site
• Other targets
• To be determined
• By whom?

• Criteria for selecting targets
• Use of the target?
• Size of the target?
• Cost of the target?
• Importance of the target?
• Easeofuse factor?
• Availability of connectors

23
CSA MultiSearch

• CSA will be doing a phased implementation
• Links between applications
• Illumina ? MultiSearch
• Search other databases
• MultiSearch ? Illumina
• Search Scholars Portal databases
• Integration of Illumina MultiSearch results
• This will enable Combined Searching

24
Combined Searching

• will help our users find their way through the
  maze of resourcesthat we offer
• and will make them happier with searching

25
Combined Searching
Single interface Combined search agent search
engine
Unified search Federated search
Many databases
Single search result
26
Mixed / Hybrid Searching
Distributed
Combined
27
Features at a glance
Complexity
Convenience
28
Transition Path
29
One Size Does Not Fit All

• The new models of searching offer our users
  convenience,
• but as effective as they are
• we must recognize that
• they are not a panacea
• they have their limitations
• some users will need native searching

30
It is all about the user

• What do we want from our users?
• or
• or do we want

31
Questions
32
Demonstration

• Disclaimer
• We are just in the early stages of
  implementation so I can show it only at a
  superficial level and may not be able to answer
  detailed questions

33
Any more questions?
34
Not embracing new solutions is todays equivalent
of burying our heads in the sand
35
(No Transcript)