22
This article was downloaded by: [Nova Southeastern University] On: 07 October 2014, At: 23:36 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK The Serials Librarian: From the Printed Page to the Digital Age Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/wser20 Metasearch in the Users' Context David Lindahl BS, MS a a University of Rochester River Campus Libraries , Rush Rhees Library, Rochester, NY, 14627, USA Published online: 12 Oct 2008. To cite this article: David Lindahl BS, MS (2007) Metasearch in the Users' Context, The Serials Librarian: From the Printed Page to the Digital Age, 51:3-4, 215-234 To link to this article: http://dx.doi.org/10.1300/J123v51n03_16 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is

Metasearch in the Users' Context

  • Upload
    david

  • View
    214

  • Download
    1

Embed Size (px)

Citation preview

This article was downloaded by: [Nova Southeastern University]On: 07 October 2014, At: 23:36Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

The Serials Librarian: From thePrinted Page to the Digital AgePublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/wser20

Metasearch in the Users'ContextDavid Lindahl BS, MS aa University of Rochester River Campus Libraries ,Rush Rhees Library, Rochester, NY, 14627, USAPublished online: 12 Oct 2008.

To cite this article: David Lindahl BS, MS (2007) Metasearch in the Users' Context, TheSerials Librarian: From the Printed Page to the Digital Age, 51:3-4, 215-234

To link to this article: http://dx.doi.org/10.1300/J123v51n03_16

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinions and viewsexpressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of theContent should not be relied upon and should be independently verified withprimary sources of information. Taylor and Francis shall not be liable for anylosses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly orindirectly in connection with, in relation to or arising out of the use of theContent.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan,sub-licensing, systematic supply, or distribution in any form to anyone is

expressly forbidden. Terms & Conditions of access and use can be found athttp://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

Metasearch in the Users’ Context

David Lindahl

ABSTRACT. The article describes the inner workings of metasearchand makes the case that there are many opportunities to improve thistechnology. Metasearch can be improved by focusing on the user’ssearch context. Examples of this are provided by the University of Roch-ester River Campus Libraries (UR) development of a UR metasearchinterface. The benefits of identifying and addressing stakeholder intereststo overcome obstacles with moving the technology forward are explor-ed. A specific framework and a mechanism to allow automatic databaseselection for metasearch users are presented. doi:10.1300/J123v51n03_16[Article copies available for a fee from The Haworth Document Delivery Ser-vice: 1-800-HAWORTH. E-mail address: <[email protected]>Website: <http://www.HaworthPress.com> © 2007 by The Haworth Press, Inc.All rights reserved.]

KEYWORDS. Metasearch, OpenURL, automatic database selection,user studies, usability

Today’s academic library users have high expectations for search en-gines. They expect to enter keywords into a box and get instant results.This expectation is difficult to meet because the library Web site con-tains many starting points that users must discover in order to be suc-

David Lindahl, BS, MS, is Director of Digital Library Initiatives at University ofRochester River Campus Libraries, Rush Rhees Library, Rochester, NY 14627. He hasalso worked as a project manager at the Xerox Palo Alto Research Center, and as asoftware engineer at Xerox Corporation. He has written several articles, and is a regularspeaker on academic library technology and user-centered design.

The author would like to thank his colleagues Nancy Foster, Susan Gibbons, JudiBriden, Jennifer Bowen, and Jeff Suszczynski at the University of Rochester, for read-ing a draft of this article and for suggesting valuable revisions.

The Serials Librarian, Vol. 51(3/4) 2007Available online at http://ser.haworthpress.com

© 2007 by The Haworth Press, Inc. All rights reserved.doi:10.1300/J123v51n03_16 215

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

cessful. One solution is to offer metasearch on the library Web site.Metasearch is software that provides a single search interface across arange of library resources.

This paper will begin by describing the inner workings of metasearchand make the case that there are many opportunities to improve thistechnology. Metasearch can be improved by focusing on the user’ssearch context, as has been done by the University of Rochester RiverCampus Libraries (UR) through the development of a UR metasearchinterface. Secondly, it is necessary to identify and address stakeholderinterests in order to overcome obstacles with moving the technologyforward. Finally, a specific framework and a mechanism to allow auto-matic database selection for metasearch users will be presented.

USABILITY ISSUES

Academic libraries today face a huge dilemma. Library Web sitesconnect patrons to a wealth of content and services, but are often diffi-cult to use. The systems incorporated into the library Web site are notwell integrated, and the content is distributed across distinct silos,which are collections of content, each with its own user interface. Silosmake content more difficult to find because users have to know where tolook. Some examples of silos include the library catalog, individualsubscription databases, and institutional repositories.

Silos create an overwhelming number of starting points that librarypatrons must learn to navigate in order to be successful. Many silos arelabeled with meaningless terms, and the silos themselves have unclearboundaries as to what is and is not available. At the University of Roch-ester, search log analysis has revealed that library users often look in thewrong place for specific kinds of content. For example, they look forcatalog content like books and journal titles in the Web site search en-gine (which searches only the Web site and not the catalog), and theylook for article content in the OPAC (which does not contain articles).

Most library Web sites offer a set of pathways to help guide users toappropriate silos for their needs. Examples of pathways to subscriptiondatabase content include lists of databases organized alphabetically andby subject. These pathways support people who recognize that theymust use databases to get to article content. The subject pathway helpspeople who have not yet learned about databases by their names. Butthis pathway requires people to be able to map their search goal into a

216 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

specific subject area. Usability studies have shown that the typical un-dergraduate cannot do this mapping reliably.1

Reference desk services, bibliographic instruction sessions, and in-formation-literacy classes teach users how to access subscription-data-base content. The problem is that students still flock to Google. Googlebrings back results without extra steps, subject choices, additional in-structions, or multiple interfaces. Google is easier to use, but it doesn’tsearch all the content in subscription databases, so it is no substitute forsearching those databases. Still, libraries need to present a consistent,seamless experience to users to make them successful with library tools.In libraries, this priority has taken a back seat to delivering precise andcomprehensive search results.

When Web site users search in the wrong silo or enter a query with-out the necessary precision, accuracy, and limits, they do not get usefulresults. But most students have the perception that they are skilledsearchers. They are always able to find useful content on Google. Whenlibrary search engines don’t return good results, user perceptions aroundtheir own search skills lead them to believe that the library does not havewhat they are looking for. In most cases, this is not true. In fact, a librar-ian could provide assistance with the search, guiding the patron to thecorrect silo, and helping to write the well-formed query that is needed.In most cases, the librarian will not get this opportunity because the stu-dent has already re-tried their search in Google and is satisfied by the“good-enough” results. Libraries have better content than Google alonecan provide, but library user interfaces lead patrons to believe the oppo-site is true. Students believe Google has more content to offer than thelibrary.

Libraries are struggling with this problem, but solving it will requirenew systems, new metadata standards, new publishing models, and newuser-interface designs. The solution will lie in refocusing on patrons,and designing interfaces that meet their expectations. Technologiesmust be hidden under a user interface that responds to what patronswant, not what the technology has to accomplish.

Library application vendors have been slow to respond, and most li-braries don’t have an established user-centered design (UCD) processin place.

User-centered design (UCD) is an approach to design that groundsthe process in information about the people who will use the prod-uct. UCD processes focus on users through the planning, design,and development of a product.2

E-Files 217

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

A UCD process would help to guide library priorities and begin tosolve some of the usability issues with current offerings. It would alsoallow libraries to articulate their needs to vendors.

Libraries face changing formats, delivery mechanisms, and publish-ing models. As new user needs emerge, libraries are forced to recon-sider their priorities. For all of these reasons, libraries must continue toshift their resources toward user-focused, Web-based offerings.

SINGLE SEARCH

Google and Amazon by their popularity and visibility set expecta-tions for how library Web sites should work. Google searches billionsof Web pages and brings back results instantly. Amazon effectivelysearches metadata on books, magazines, and even toys with a singlesearch box. Users don’t expect a Web site to have more than one kind ofsearch box for different types of content. Libraries that attempt to ad-dress this primarily through increasing the “information literacy” oftheir users are missing a huge opportunity. A better solution is to meetthe expectation with a single search box across all scholarly content.Metasearch technology enables libraries to offer this search box.

Metasearch is a technology that allows a user to search distinct data-bases concurrently, retrieve the results, and view them in a singlemerged result list with identifying metadata for users. A link on each re-sult then points users toward full-text services for items they select. Allof this can be done in a single user interface. This is in contrast to thevarying native database interfaces that users typically need to deal withwhen searching for full-text articles. Metasearch is becoming popularamong large university libraries that subscribe to multiple databases. Inorder to simplify and unify access, metasearch can be offered as a singlesearch across all databases.

Metasearch is not a single technology. It is made up of a number oftechnologies that are distributed on the internet. A library Web site, meta-search server, OpenURL resolver, subscription databases and othercontent repositories must all work together to deliver the single searchexperience. Conceptually, it is useful to separate this technical infra-structure from the interface. It is possible to create a seamless user inter-face, which hides all of these disparate pieces of technology from theuser.

218 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

From the library systems perspective, a metasearch system brings to-gether two technologies for getting patrons to database content. Thefirst technology is the metasearch engine. This software searches data-bases and brings back citation metadata. The second technology is theOpenURL link resolver. This software allows library patrons to movefrom citations in one database to full-text content in another. Thesetechnologies work together to create a pathway that begins with the userentering query terms and ends with the delivery of full-text content.

The metasearch engine begins by collecting query terms and data-base choices from a user and then submitting the query to each remotedatabase site simultaneously. The software then gathers the results fromeach database and displays a merged results list that is sorted and de-duped. Each item on the results list may include citation metadata, anabstract, and a link to full-text services.

The OpenURL link resolver enables patrons to move from the meta-search results screen to full-text items. This capability typically appearsto library patrons as a graphical button on the item-record screen or nextto each item on the search-results screen. When clicked, the user istaken to a menu with a set of services for the cited content. These ser-vices include options for obtaining full text from a database, print hold-ings from the library catalog and interlibrary loan services. Someresolvers bypass the menu and take patrons directly to the full text.

Libraries have many options when selecting metasearch engine andOpenURL link resolver software. Because metasearch depends on bothtechnologies to work, vendors often bundle the two products together.Depending on the specific needs of a library, it is reasonable to purchase ametasearch engine and OpenURL link resolver from two different ven-dors and configure them to work together. Some examples of metasearchengines include Metalib from Ex Libris3 and ENCompass for ResourceAccess from Endeavor.4 OpenURL link resolver examples include SFXfrom Ex Libris5 and LinkFinderPlus from Endeavor.6

Typical metasearch products come with an out-of-the-box user inter-face that can be customized to fit the look and feel of a library Web site.Some metasearch products offer a Web services interface to bypass theout-of-the-box interface entirely. Web service interfaces allow librarieswith a programmer on staff to design their own user interfaces fromscratch or to integrate metasearch into their library Web site with a greatdeal of flexibility. When combined with a UCD process, this is an idealapproach to meeting patron needs and expectations.

E-Files 219

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

METASEARCH AND USABILITY ISSUES

The promise of metasearch is that it will be a single search box acrossall content that the library has to offer. For many patrons, metasearcheliminates the need to learn multiple native user interfaces and the needto bounce around among them. It lets patrons identify and select itemsfrom relevant databases on a single results screen. It even has the abilityto deliver Web content indexed by Google in the same results set.

Metasearch has the potential to offer a “single search box across alllibrary resources” similar to what Google does for the Web. Of course,promises and reality do not always match. There are many issues withmetasearch, and some of these issues are unacceptable to users. Why aresome libraries interested in a new technology that makes unacceptabletradeoffs? The answer lies in where we were before metasearch waspossible. Metasearch meets the needs of users who would most likelyhave bypassed library resources entirely by going to Google instead ofthe library. It allows them to discover and evaluate scholarly content se-lected by librarians and can bring them back.

Metasearch and OpenURL are amazing enabling technologies, butuser needs have not been fully incorporated into the user interfaces pro-vided by vendors. Typical out-of-the-box interfaces on commercialproducts are complex, have too many screens, and closely follow thesteps the system needs to go through to perform the search. In contrast, ausable interface would hide all steps except what the user expects orneeds to see. Thus, the first round of out-of-the-box interfaces have anumber of usability issues, which fall into four main categories: speci-fying queries, selecting databases, identifying and selecting results, andacquiring items. The specific issues, and what is being done in the meta-search community to address them, are discussed below.

Category 1: Specifying Queries

For many users, specifying queries is easier with metasearch thanwith individual native database interfaces. With metasearch, patronsneed to learn only one interface in order to search many databases.Some patrons would also like the ability to specify a search type, such asan author search or a title search, and would like to set specific limitssuch as a date range. This level of specificity is often impossible acrossa range of databases, since each database formats the metadata differ-ently and uses a different access mechanism. In this case, the results aretypically not as good as what the native interface returns. Efforts of the

220 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

NISO Metasearch Initiative7 should help with this issue by defining aset of query and metadata standards that will guide stakeholders in theirfuture development decisions.

Category 2: Selecting Databases

Metasearch engines do nothing to assist the user in knowing whichdatabases are the best to search. There are two common ways that librar-ies have configured the user-interface layer on top of the metasearch en-gine to help with this problem. The first method is to group databases bysubject and ask patrons to select a subject area along with their queryterms. The second is to present a search box without any other choicesto make. This method hides the database selection process entirely, anduses a small set of pre-selected, general databases. In both cases, the da-tabase selection process is not automated but instead it relies on a user tomake a choice, or on a pre-selected, best-guess generalization about thedomain of the search. A better solution would be to have the system au-tomatically select databases based on query terms. This solution doesnot seem to be forthcoming from commercial metasearch vendors.

Category 3: Identifying and Selecting Results

Identifying and selecting results is efficient with metasearch becausepatrons can do it across multiple databases on a single screen. The prob-lems here depend on specific database and metasearch vendor imple-mentations. Inconsistencies across databases mean that some resultswill be presented without abstracts and others without the key metadataelements necessary for patrons to identify their desired results. Under-graduates want the instant gratification of electronic full-text, but full-text availability information is almost never displayed on the resultsscreen. Other issues include poor relevancy-ranking algorithms, over-whelming numbers of results, slow application-response times, and da-tabase failures causing patrons to miss out on a portion of the results.These issues are gradually being addressed in new metasearch productsand by individual library implementations.

Category 4: Acquiring Items

OpenURL link resolver menus are used to deliver full-text to meta-search users. But when undergraduate students see an OpenURL linkmenu instead of the full-text they expected, they often assume that thesystem didn’t work and they give up. OpenURL menus can lead to long

E-Files 221

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

click-paths and backtracking to determine whether items are available atall, either in electronic or in print formats. There are often multiple pop-up windows, and users must move between them to manually copy andpaste citation information as they re-submit their searches. The usabilityproblems with acquiring items have been solved at the University ofRochester with a homegrown application called GUF (Getting Users toFull- text). The GUF application is described in the next section of thispaper.

All of these technology issues and design tradeoffs have led to somevery real resistance to using the technology. Metasearch technology canobscure some library content, bring back the wrong content in the wrongorder, make professional librarians distrust how queries are applied in da-tabases, and generally return unacceptable results. This results in reluc-tance to use metasearch by librarians and other stakeholders.

Despite these issues, metasearch is an improvement over searchingnative database interfaces. Even current metasearch offerings succeedat delivering database content to students who would otherwise besearching Google. But the technology will need sustained work to evolveinto something that functions well enough for all users. Metasearch of-fers the most promise, of any existing technology, to make academic li-braries relevant to patrons in the Web context.

METASEARCH INTERFACE WORKAT THE UNIVERSITY OF ROCHESTER

The University of Rochester River Campus Libraries has been offeringusers metasearch since the summer of 2003. We are currently in ourfourth-generation interface design, although we have been making incre-mental improvements since we first released this feature on our Web site.The front-end interface to our metasearch offering is called Find Articles.Our goal has been to provide a pathway for finding article content thatmeets the needs of undergraduates. Our focus has been to deliver a fewgood articles to students quickly and easily. Over the course of our user-interface development, we have uncovered and addressed numerous us-ability issues and I will discuss some examples in this paper.

Interface Redesign

The out-of-the-box interface that came with our metasearch engine(ENCompass for Resource Access from Endeavor) required the user to

222 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

click at least six times to get to full-text content. We eliminated stepsfrom the process by making choices automatically for the user. Our li-brarians pre-selected a set of general-article databases to be searched,and our programmer wrote a script that merged the results from the dif-ferent databases without the user having to click a button to merge them.We designed an interface that gets the patron from search box to full textin two clicks. This is the same number of clicks patrons must makewhen starting from the Google homepage to get to Web content.

GUF

The out-of-the-box interface that came with our OpenURL link re-solver (SFX from Ex Libris) presented the patron with a list of servicesavailable for a given citation. In our usability testing with undergradu-ates, this menu was found to be confusing. These patrons expected tosee the full text, not a menu of services. Our programmer, Jeff Suszczy-nski, created new software called GUF (Get Users to Full text) to bypassthe OpenURL resolver (SFX) menu and take users directly to the re-source. The GUF software depends on the Web service interface that isincluded with the SFX product in order to function. GUF works by fol-lowing links, pre-fetching pages, and performing error checking on be-half of the patron. All the patron sees is an animation while GUF findsthe best-available version of the item according to pre-programmed pri-orities. Online full text is the highest priority, followed by a call numberand location information for a print copy in the library stacks, and an op-tion to order from interlibrary loan is presented as a last resort. GUF ishelping us to succeed at getting novice users to article content beforethey give up and turn to Google.

Improved Results Set Display

In our usability testing, we noticed that students were getting resultsfrom their keyword searches but did not understand why certain articleswere on the results page. When the vendor of our metasearch servermade it available, we added partial abstracts to the results page to helpthe patrons identify the results, but this was not sufficient. In our secondattempt, we added query-term highlighting in titles and abstracts on theresults page. This approach proved successful in subsequent rounds ofusability testing. The addition of the keyword-highlighting feature wasmade possible through the XML and XSLT user-interface layer that our

E-Files 223

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

metasearch vendor (Endeavor) offers on top of its product. Our pro-grammer implemented the keyword-highlighting routine in an XSLTstyle sheet.

Expectation of Relevancy Ranking

The next issue that we dealt with was based on a user expectation anda missing piece of technology. Our users expect results to come back inrelevance order, but our metasearch software was unable to do rele-vance sorting. We therefore chose to return results in alphabetical orderas the best-possible solution available to us. However, the relevancy ex-pectation of users is so strong that in usability testing the users did notnotice that the results were sorted alphabetically. This was a huge prob-lem that led users to consider only the first few hits when they actuallyneeded to review them all in order to find the best hits in the result set.We needed to make it clear that the results were in alphabetical orderand we needed to keep the result set short. We were successful in de-signing a solution, but this was merely, to quote Andrew Pace, “lipstickon a pig.” Truly solving this issue will require our metasearch vendor toprovide relevance ranked results.

Course Clusters

Usability tests at the University of Rochester have shown that stu-dents are course-oriented and that they often have difficulty selecting asubject for their search. In response to this, we created “course clusters”which are collections of databases selected by bibliographers that arerelevant to specific courses. Like many library sites, our library offerscourse pages with links to library resources and reserve materials. Librarypatrons at the University of Rochester can start a metasearch from theircourse-specific Web pages. They are offered a search box that searchesdatabases relevant to the course (see Figure 1). This allows a well-tar-geted database search for students without their having to choose adatabase or subject. This is a significant improvement over a list of data-bases on a subject guide.

Pathways to Content

At the University of Rochester, we have implemented novice and ex-pert pathways on our Web site to get users to subscription database con-tent. Novice pathways address the needs of new undergraduates. Expert

224 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

pathways are designed for users who have some experience searchingdatabases. Pathways help solve some of the problems that metasearchintroduces. Maintaining existing pathways ensures that experienced us-

E-Files 225

FIGURE 1. Metasearch from a Course-Specific Web Page

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

ers continue with whatever level of success they had before, but the ad-dition of new pathways opens up subscription database content to a newuniverse of patrons and may give experienced users some new tools andunderstandings as well.

STAKEHOLDER INTERESTS AND USER NEEDS

There are many stakeholders involved with metasearch technology.These include librarians, patrons, publishers, database vendors, and meta-search vendors. Each stakeholder has a unique set of interests in howscholarly content is made available. The key to the success of meta-search lies in the continued involvement of the stakeholders, and theevolution of metasearch technology to best meet user needs.

It is sometimes difficult to separate stakeholder interests from theirpositions. Each stakeholder has an interest on how metasearch technol-ogy should be implemented. For database vendors there is an economicinterest. For librarians there is a professional interest. Many stake-holders also have positions that they are not willing to move away from,even though there may be better solutions that meet multiple stake-holder needs. One of the positions that database vendors take is that theydo not wish to compromise their brands by having metasearch softwarehide their logos and proprietary interfaces. For librarians who wantcomprehensive and precise results sets, the position is that metasearchserves to dumb-down the experience and therefore shouldn’t be used. Inorder to make metasearch work well, it is critical to separate positionsfrom interests. A focus on interests alone can lead to innovative solu-tions that meet multiple stakeholder needs. Database vendors and librar-ians provide two good examples.

Database vendors don’t want additional load on their servers fromindiscriminate metasearch sessions. They only want their servers to re-spond to appropriate searches that truly target the content in their subject-specific databases. Their position has been to demand user agent identi-fication. Currently, database vendors have no way to distinguish be-tween library users that come in through the native interface and thosethat metasearch. User agent identification would require the metasearchengine to identify itself as such. Instead of focusing on this partial solu-tion, new technologies that target appropriate databases might betterserve their interests and serve librarian interests as well.

Librarians dislike the imprecise nature of metasearch, which tends toaccept mostly “keyword-anywhere” searches and brings back only a

226 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

few good results. This is in stark contrast to the precise and comprehen-sive results that are retrieved from a well-executed native interfacesearch on an individual database. However, when library users partici-pate in usability tests with library interfaces, they often fail to find arti-cle content at all. As a result, instead of getting a few good results from adatabase selected by a librarian, they are turning to content from theWeb that has not been vetted by librarians at all. Librarian interests areserved best by working to provide a metasearch pathway and workingto improve the precision of metasearch technology, not by trying todiscredit it.

If the future of metasearch is driven by user needs, then stakeholder in-terests will be served. Scholarly publishing is being shaped by some ofthe same forces that have changed music publishing. Electronic deliveryof music, combined with community-based tools for sharing content andcommunicating about it, led to a complete overhaul of the economics andmechanics of music publishing. Apple, along with its iTunes service, hasbecome one of the largest beneficiaries of this new publishing model. Thelesson here is that in a Web-based world, user needs will determine thebest approach to content delivery. If we successfully address user needsin academic libraries, users will flock to our solution.

Making metasearch technology successful involves getting all of thestakeholders on-board. By identifying and responding to the needs ofeach of these groups, metasearch will evolve into a successful tool thatexceeds current expectations for scholarly content search tools. Meta-search technology is not the only thing that has to change. Scholarly pub-lishers, librarians, aggregators, database vendors, metasearch technologyvendors, and library administrators all have to reevaluate their priorities,change their practices, invest in new approaches, and focus on the end user.

The first step toward meeting stakeholder needs is to identify them.The emerging metasearch community has become aware of many ofthese needs and has started to address them. We need to continue to ap-ply appropriate methodologies to uncover the complete set of needs.This will create an opportunity for the community to work towards solu-tions that meet user needs while preserving stakeholder interests. In thisway, stakeholders will commit resources to the development of the so-lutions. Below is a summary of stakeholder needs and interests derivedfrom several studies conducted at the University of Rochester RiverCampus Libraries. These interests come from a study of faculty needssponsored by the Institute for Museum and Library Services,8 a partiallycompleted study of Undergraduates and their paper-writing process,and ongoing usability studies.

E-Files 227

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

UNDERGRADUATE INTERESTS

• Complete course requirements quickly and easily• Choose paper topics• Find a few good articles for a paper• Find general content in some cases and specific content in others• Create a bibliography• Think in terms of courses, not subject areas• Write• Work anywhere, inside the library or away from it• Save full-text articles and citations on a personal computer or drive• Print articles• E-mail articles to allow others to edit their paper• Learn as few interfaces as possible• Have a Google experience

AGGREGATOR, PUBLISHER,DATABASE VENDOR INTERESTS

• Lower the technology barrier of entry for database vendors• Preserve past investment in native user interfaces• Preserve investment in brand recognition that metasearch can ob-

scure and hence threaten• Avoid technology expenses benefiting only a few customers–the

unproven revenue model• User agent identification• Reduce impact on content provider systems• Understand how they fit into the future publishing models• Avoid having their records de-duped in favor of hits from another

database• Understand how hits from their database are ranked in relation to

hits from other vendor databases• Use standards to help guide their investment• Use cost-effective technologies• Use open-source software as appropriate• Preserve successful revenue streams and seek out new ones

LIBRARIES AND LIBRARIANS

• Facilitate librarian-to-user communication• Conduct precise and comprehensive searches

228 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

• Meet user needs• Deliver services that distinguish their offerings from Google• Preserve the Reference interview• Participate in design of metasearch user interface, choices, trade-

offs• Participate in the selection of content that can be metasearched• Use metasearch software that is easy to configure and maintain• Use reliable software• Maintain capability that comes with existing tools• Make resources readily accessible to users• Make searching easier• Expose resources that might otherwise be hidden and go unused• Get staff consensus when moving forward with a new digital tool• Avoid dumbing-down the interface

FACULTY

• Keep up in their field• Avoid being any busier• Know that they are searching everything• Make their own work easily findable and increase citation rates• Use tools that are precise and comprehensive.• Get access to the latest published information on their topic

AUTOMATIC DATABASE SELECTION

Most metasearch engines make no attempt to automatically selectappropriate databases for patrons. Libraries that implement metasearchmust decide whether to select databases for the user, offer subject group-ings of databases, or just let the user select databases directly. Most li-braries offer one or more of these pathways. The goal is to deliverprecise and comprehensive results from an easy-to-use search interfacewhile targeting only the most appropriate databases. Without automaticdatabase selection and a new kind of user interface, the only option is tomake tradeoffs on ease-of-use, precision, comprehensiveness, and loadon database vendor servers.

Here I will propose a framework and a mechanism for automatic da-tabase selection that is closely tied in with meeting user needs. The goalis to return a results set that is both precise and comprehensive across all

E-Files 229

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

available databases. The system should not require the user to select asubject nor should it waste database-vendor resources by searching allof them indiscriminately. This will be achieved through a set of Webservices offered by database vendors and an interface that collects pa-tron input at various stages in the process.

Most users enter search queries that are too imprecise for an auto-mated database selection system to work well. The first step is to designa user interface that uncovers the patron’s desired search context. Bycontext, I am referring to whatever additional information is needed be-yond the query terms that the patron enters, in order to deliver preciseand comprehensive results to the user. Examples of context include sub-ject area, discipline, author, and time period.

Search queries can have different meanings in different contexts.Providing useful results depends on the system’s ability to determinethe desired context from the user. For example, if a patron enters a per-son’s name, they might want materials written by that person or materi-als with that person as a subject. Most discipline-specific databaseshave a thesaurus that assists users in understanding the terminology ofthe discipline. The problem is that a term entered into a metasearch en-gine without a specified discipline can’t be applied to the correct thesau-rus. Many terms have different meanings in the context of differentdisciplines. Without a better understanding of what exactly a user islooking for, precise automatic database selection is impossible.

When users begin their search, they can’t necessarily identify thecontext that they are interested in. Even when they can, there is no goodway for them to express it. Often, they cannot identify the context wellenough to express it in their keywords. Usually, the keywords that auser types can fall into more than one context, so in order to bring backappropriate results, the system needs to gather context information fromthe user. The best way for a system to do this, I propose, is iteratively,while the user is looking at results. The actions associated with navigat-ing the results should be used to determine context. The context canthen be used to refine the selection of databases and the results withineach database. Successive iterations could change the mix of databasesbeing searched as more results are returned in the desired context.

What if you could ask each database how relevant it is for a givenquery in order to select appropriate databases for your metasearch? Ifthis process were fast and accurate, what might it look like? In order tobe successful, the initial entry of keywords into the search engine andsubsequent initial display of hits should not make any assumptionsabout context. In fact, all available sources should be queried. This radi-

230 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

cal approach to library metasearch can work, but only if new Web ser-vices are provided at the database end.

Current metasearch engines can only interact with databases in oneway: by sending a query to a database and getting back some results. Wedo not want metasearch engines to interact with all available databasesin this way, both because this would produce tremendous load on ven-dors’ servers, and because the majority of the results would not be rele-vant to the given query terms. The alternative to this indiscriminatebroadcast metasearch is to target the best databases for a given search.But how can this be done?

Implementing new, metasearch-specific Web services at the data-base end can help with this problem. These new services can be light-weight and would need to deliver a much smaller amount of data to themetasearch engine. For example, a service that would return the appro-priateness of a database for a particular query could do so with a smallamount of statistical data. These services can also be used to determinethe range of contexts that the users’ search terms fit into and then revealdatabase coverage for the combination of context and keywords.

To make this work, we need a Web services infrastructure for select-ing databases based on a comparison of query terms against the full-textindex of each database. The results of this selection step should be sta-tistical information about the match between the query and the databasecontents. A second Web service might provide a set of contexts inwhich the query makes sense in the database being searched. In thiscase, context refers to the relationships that exist between all of theitems in the theoretical “ideal” result set for the users’ needs. One exam-ple of context might be to return subject headings for a given set ofquery terms. Another example would be to return identified relation-ships that exist between items in a result set. The Web service would re-turn the values of specific fields that match across two or more items ina result set for a given query.

In this scenario with a new set of Web services, the user would be in-teracting with a much richer display. The items in the display would re-veal the range of contexts that the query terms fit into. Users would seeresults grouped by their relationships with each other. The item meta-data would be there to help users distinguish one item from another.Metadata common to items in a grouping would only display once forthe group. Users could then identify desired contexts, which would beused behind-the-scenes for refinement of results.

Many database vendors tout a thesaurus capability as a feature oftheir native interface. In the metasearch environment, thesaurus ser-

E-Files 231

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

vices should be delivered to the metasearch engine. When a thesauruslookup is used for a particular search query, the database should revealthis to the metasearch engine. In addition, thesaurus lookups should bemade possible with a Web service call to the database.

A Web service infrastructure for database selection will require theparticipation of database vendors, metasearch engine vendors, and li-braries. Each of these stakeholders will need to make an investment inorder to advance in this area and move the burden of database selectioninto the realm of automation. As we move toward better metasearch, theNISO Metasearch Initiative can help by defining standards for the data-base-selection Web services that are needed.

METASEARCH AND LIBRARIANSHIP

As metasearch technology evolves, it will serve as a platform for li-brary innovation. Libraries can begin to tackle new opportunities for li-brarianship in the context of metasearch. The following is a list of someof the opportunities for library-sponsored web applications that makesense in a metasearch environment:

• Provide online evaluation of content from librarians• Offer alternative editions of a work using the OCLC × ISBN ser-

vice• Offer a way to view other works by the same author• Offer a way to view other works on the same topic• Connect patrons to related blog content• Present evaluative information (created by librarians, researchers,

peers, etc.) including the authoritative nature of the work, andonline reviews

• Provide tools to help create a bibliography• Provide tools to support annotation and sharing of the work• Support personal collections• Provide tools to receive and evaluate results at a later time• Build pathways between library content discovered through meta-

search, and appliances like the laptop, iPod, and Tivo• Support discussion with colleagues, classmates, instructors, and

librarians

Many librarians consider the reference interview to be a criticalmethod of their discipline. Virtual or chat-based reference service is be-

232 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

coming a common offering on academic library Web sites. Perhaps vir-tual reference service could be combined with metasearch in a new way.Patrons could contact a reference librarian while using metasearch soft-ware by means of an instant messaging or live voice-chat service. Thekey to making this work would be to deliver the search results to boththe patron and the librarian simultaneously. Web-based collaborationproducts make this a realistic possibility today, assuming the meta-search engine is delivered as a Web service.

Because so many aspects of librarianship are changing in the contextof the web, it is wise to rethink how libraries should be organized andhow to best apply library resources to Web librarianship. Library con-tent is increasingly provided in digital formats, and user expectationscontinue to change. Librarianship includes the evaluation, acquisition,and preservation of resources. It also involves providing access to re-sources and making them findable and usable. The Web is a great op-portunity for libraries to do this in new ways. Metasearch can facilitatethe participation of all libraries.

Given the range of stakeholder interests and the distributed nature ofmetasearch technology, the future of metasearch technology is difficult topredict. However, if we move down the path of Web services and stan-dards, metasearch will become a stable platform for library innovation.NISO standards will guide implementation efforts and open-source soft-ware will play a role in making software available to all libraries. Themetasearch user interface could evolve in many different ways, but if li-braries focus their efforts on UCD, they will make progress toward thegoal of having a single search interface across all library content that isprecise, comprehensive, and usable for all library patrons.

REFERENCES

1. B. Reeb and S. Gibbons, “Students, Librarians, and Subject Guides: Improving aPoor Rate of Return,” portal: Libraries and the Academy 4, no. 1 (2004): 123–130.

2. Usability Professionals’ Association, “Resources: About Usability: What is User-Centered Design?,” http://www.upassoc.org/usability_resources/about_usability/ what_is_ucd.html (Accessed December 2, 2005).

3. Ex Libris LTD, “Ex Libris - Metalib–Overview,” http://www.exlibrisgroup. com/metalib.htm (Accessed December 4, 2005).

4. Endeavor Information Systems, “ENCompass for Resource Access,” http://en-compass.endinfosys.com/era.htm (Accessed December 4, 2005).

5. Ex Libris LTD, “Ex Libris – SFX – Overview,” http://www.exlibrisgroup.com/sfx.htm (Accessed December 4, 2005).

E-Files 233

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014

6. Endeavor Information Systems, “LinkFinderPlus,” http://www.endinfosys.com/prods/linkfinderplus.htm (Accessed December 4, 2005).

7. National Information Standards Organization (NISO), “NISO Metasearch Ini-tiative,” http://www.niso.org/committees/MS_initiative.html (Accessed December 4,2005).

8. N. Foster and S.Gibbons, “Understanding Faculty to Improve Content Recruit-ment for Institutional Repositories,” D-Lib Magazine 11, no. 1 (January 2005).

doi:10.1300/J123v51n03_16

234 THE SERIALS LIBRARIAN

Dow

nloa

ded

by [

Nov

a So

uthe

aste

rn U

nive

rsity

] at

23:

36 0

7 O

ctob

er 2

014