driving discovery: do you have the keys to fair linking? (it’s about knowledge and library...

Post on 05-Jul-2015

182 Views

Category:

Education

3 Downloads

Preview:

Click to see full reader

DESCRIPTION

2014 Charleston Conference Presentation by Scott Bernier, EBSCO Charleston Neapolitan Session Thursday, Nov 6, 10:30 AM

TRANSCRIPT

Charleston Conference

November, 2014

Fair Linking &

Library Choice

A discussion of custom full-text link set up

What makes linking “Fair”?

the discovery vendor steps aside and

gives libraries complete control over

the full text links that appear in their

result list – when and how

In simplest terms

Hence, the title of this session….

• But how do you adjust the seats?

• Which radio station do you tune into?

• At what temperature do you set the

climate control?

Libraries are driving: You have the

controls & keys [to the car]…

Best Practices?

…but if we collectively understand the way

the services work, we are in a position to

consider the options and move toward an

ideal implementation (for each library)

Is there a driver’s manual for all of this?

It may not be that one-size-fits-all….

Perception vs. Reality

Discovery Bias

& Fair Linking

A conversation that brought

me here

A 2-Part Equation

These 2 components are completely

de-coupled in the EDS experience.

1.Records appearing in the result list

(Relevance Ranking)

2.The full text associated with a given record

(Linking Set-up)

Relevance Ranking Results

The Search Team at EBSCO really has one

goal as it relates to EDS relevancy ranking:

Ensure that the best possible

results for every search

query appear at the top of the

result list every time.

Relevance Ranking Results

The source/provider of a record is not a

factor in the ranking algorithms

But, the level of data

available on which to base

ranking decisions is very

important.

EDS Relevance Ranking Ingredients

1. Matching word frequency

2. Metadata field weighting

3. Value ranking

4. Exact field match boost

5. Local collection weighting

No simplistic

formula for

relevance ranking–

multiple factors

blend to deliver

relevant results.

Metadata Field Weighting

1. Match on subject headings from controlled vocabularies

2. Match on article titles

3. Match on author keywords

4. Match on keywords within abstracts

5. Match on keywords within full text

Some metadata fields count more than others for scoring.

There is NO weighting differentiation for content source/provider

More fields than

these are used for

field weighting.

Each available data field heightens

knowledge about an article, which contributes

to the ranking potential of an article.

There is only benefit to the content partner

(and ultimately the end user) when EDS has

more information about their titles (subject

indexing, full text, abstracts, etc.)

EDS aims to improve the user’s experience

with the library and its resources

usability studies with undergraduate

students, graduate students and faculty lead

us to include a “value” ranking based on user

expectations and preferences.

Usability Studies

Value Ranking

1. Publication date

2. Publication type

3. Peer reviewed or not

4. Document length

Specific content attributes of matching records contribute to

relevance scoring.

More attributes than

these are used for

value ranking.

EDS

Full-Text Links

Library Choice

Full Text Links in EDS: Custom Links

Each customer controls these custom

links in the following ways:

1. Which full text they wish to display

2. The order in which their links may appear

(their full text preference)

3. Whether multiple links to the same full text

displays (simultaneously)

Multiple Links from Same Record?

Full Text Links in EDS

Because customers choose their links and

link order preferences, the provider/

source of a record has no bearing on

which link may appear within that record.

Sample Record from Business Source

Sample Record from Business Source

Sample Record from Business Source

Custom Links “Default Setting”

• There are no default settings

• We first must enter a library’s collection, then

they can determine their link preferences

• If we did include certain collections by default,

users could easily run into unintended pay-walls

– This is because we are likely to include resources the

library does not own

– What if we knew of customer collections in advance?

(This could have multiple benefits)

0 ACM Digital Library (EDS)

0 American Chemical Society (EDS)

0 American Institute of Physics (EDS)

0 American Physical Society (EDS)

0 Annual Reviews (EDS)

0 BioMed Central/Chemistry Central/SpringerOpen EDS)

0 Cambridge University Press (EDS)

0 Elsevier ScienceDirect (EDS)

0 IEEE Xplore (EDS)

0 Institute of Physics (EDS)

0 Lippincott, Williams & Wilkins (EDS)

0 MIT Press (EDS)

0 Nature Publishing (EDS)

0 Oxford University Press (EDS)

0 Project Muse (EDS)

0 Royal Society of Chemistry (EDS)

0 S. Karger AG (EDS)

0 Sage (EDS)

0 SpringerLink (EDS)

“Digging into the Data: Exposing

the Causes of Resolver Failure”Cindi Trainor, Eastern Kentucky University

and Jason Price , Claremont Colleges

• Link resolvers “fail nearly

a third of the time”

• Even when the link resolver

does not fail, it requires

multiple clicks to get

to the full text

Opportunities – Link Resolver

Determine Best Practices of when & how

to use or by-pass the link resolver:

• What does the user tell us?

• Consistency. Familiarity. Simplicity.

• In the user experience – is it a link that says “full

text”, or a custom branded name for the link resolver

that takes us to a menu with multiple options?

SmartLinks Plus

• What are these are how are they set up?

– Automated direct links to publisher e-journal

– For customer who purchase

journals/packages from EBSCO

• Upcoming changes to SmartLinks Plus –

to follow the rules of custom links

JSTOR Specific Links - Example

• There are 6,000+ sites using EDS

• In order to look at the full text link set-up

for a customer, we must look at the “profile

level” for each customer

• Across traditional academic sites only, we

have 7,978 current EDS profiles

JSTOR Specific Links

Of the 7,978 academic EDS profiles:

• 2,764 (34%)

have NO JSTOR LINKS AT ALL

• 3,530 (44%)

have only the “free” JSTOR link

Opportunities

An understanding of a content provider’s

customers, would:

• Provide potential to turn on links “out of the gate”

(“Defaults” without the pay-wall concern)

• Allow us to specifically identify any existing

customer that as of today has NO link turned on

in EDS for content to which they subscribe

Opportunities

We can look closely at customer link profiles

An opportunity for libraries to fine-tune

the rank/order/inclusion of links – as well

as how these links display

Opportunities for Content Providers

• Providing Full Text Data for Searching?

• eBooks Chapter Level Data?

• Documentation Improvements

– Quick Reference Guides

• EDS “how to” video on full-text link set up

Ensuring Customers & EDS Partners Are

Aware of Linking Options/Opportunities

Other Opportunities

• “The Catalog Factor”

• With discovery, the Catalog is now “competing”

with journals in a direct way

• Libraries still ‘favor’ catalog records over other

records?

• Ability to adjust catalog weighting in discovery

services – best practices

The Great News…

• We all want the same thing

• More usage & value of library resources

• Exposure of the ENTIRE collection

• Enhanced perception of the library

• But ultimately…satisfied end users who

return to the library (and tell their friends)

– Because of expedient, fruitful experience

Perception vs. Reality& Now What?

• Overarching goal is to better communicate

• Know the facts (how it works)

• Eliminate the [false] perception of bias

• Move forward

– Optimize (the user experience)

– Improve collaboration (we all want the same thing)

– Best Practices

Scott Bernier

sbernier@ebsco.com

Thank You

top related