Discovery services sift through expert resources

Internet search engines, whatever we think of them, have changed user expectations of searching for information. Expert databases and other information services and platforms are now held up against the new ‘gold standard’ of Google, irrespective of the relative quality of the information unearthed. The message from users seems clear: they like a simple search box that enables them to search ‘everywhere’ at once.

This approach challenges the traditional scholarly information approach where high-quality information might be stored in different silos, behind different interfaces and access constraints. Studies of user behaviour by OCLC and others have shown that users, especially undergraduates, are drifting away from the library’s services and moving to other sources of information, primarily web search engines.

‘The systems found in a typical library, which were developed over time around librarians’ needs to manage materials of different types, cannot be modified and brought together in a way that will provide a more productive discovery experience for today’s users,’ explains Tamar Sadeh, director of marketing at Ex Libris.

‘There’s an increasing body of research that shows that end-users find library collections confusing. They don’t know where to start and they often end up in dead-ends. Their frustration drives them back to the open web, where they feel confident in managing their own information search,’ agreed Michael Gersch, ProQuest vice president and general manager of Serials Solutions. ‘End-users want the library to be just as simple as open-web search engines. Librarians want to deliver that kind of service, too, and have been calling for single point access that unifies content.’

Federated searching, which enables searching across information silos, was a first step towards this goal. However, in large and complex collections it cannot deliver the fast, accurate results that libraries demand.

‘We saw Google as very fast, but incomplete. Conversely, federated search services (or metasearch engines) have serious speed and performance issues, as well as normalisation problems,’ explained Sam Brooks, senior vice president, sales and marketing for EBSCO Publishing.

The rise of discovery

These challenges have given rise to a new breed of search tools known as discovery services. ‘All metadata in EBSCO Discovery Service (EDS) is loaded on EBSCO servers, so there is no issue with performance of third-party servers. The only time any user leaves EBSCO servers is to gain full text via a link resolver, or to incorporate one of the minority of resources that the library has chosen to access via Optional Integrated Federation – which is much more effective than standard federation,’ pointed out Brooks.

Serials Solutions’ discovery service is called Summon: ‘The research behind the Summon service told us a discovery service needs to offer a Google-like experience at the minimum. So we built Summon to meet those expectations: it’s open – no barriers, no sign-in to starting, it’s easy to use, with no need for familiarity with a specific library database interface… just one single searchbox across the entire collection. And it’s fast: all results are returned in less than a second, with no stragglers,’ said Gersch.

‘Improving discovery means improving the usefulness of the library, enabling it to leverage its single most important competitive advantage: its collection,’ Gersch continued. ‘The challenge has been in developing technology that could support rich, user-centred discovery. The research behind Summon showed that the only way to meet user expectations in a complex library environment was to start with a completely new technology architecture: one with a unified, comprehensive index at its core to ensure content neutrality, supported by rich metadata, and built on an open source that enables it to be customised and grow.’

Sadeh of Ex Libris, which has the Primo platform, agreed on the benefits of discovery: ‘Discovery tools aim to provide a library-branded “Google experience” within the academic context in a manner that suits the institution’s mission and policies while maintaining the quality of the content and the accuracy of the search,’ she said.

‘Library discovery tools have started playing, in the scholarly world, the role that web search engines played 15 years ago in the general information-searching arena. Discovery tools differ from the previous generation of library search tools by providing an easy, quick, and unified access to information that used to be available through multiple systems and through multiple interfaces. In addition, they offer post-search options for narrowing down result lists, “Did You Mean” suggestions, and services such as RSS feeds,’ she added.

Availability of content

The content available through discovery tools typically spans over several ‘silos’ of content. This might be a library catalogue, local digital repositories, and subscribed collections. However, the scope of the content differs from one discovery solution to another, as does the degree of integration with other systems in the library.

‘We work with all information providers that are interested in exposing their collections – primary or secondary scholarly materials of all kinds – to more researchers. Obviously, more exposure triggers more usage, which is tracked by libraries and serves as input for collection development decisions,’ said Sadeh.

‘We join the call of libraries that, through the International Coalition of Library Consortia (ICOLC)’s statement, are encouraging publishers to allow their content to be made available through discovery services on a non-exclusive basis. We see the situation today as being somewhat reminiscent of the early days of the OpenURL standard; many providers were very quick to embrace it but some were more cautious; however, very soon the OpenURL was universally adopted.’

Michael Gersch ProQuest/Serials Solutions

In addition to making content available, discovery tool providers can still have a challenge to address in terms of librarians’ acceptance of discovery tools. ‘Whereas some librarians welcome the advent of such tools, others feel that the ease of search makes the information-seeking practices of users shallower. We strive to make the life of users as simple as possible, but not too simple so that librarians can feel confident that users are able to still find the information that is most appropriate for them,’ Sadeh explained.

Brooks of EBSCO also sees challenges for libraries and their users in realising that different discovery services do different things: ‘A discovery tool that doesn’t give good depth as well as breadth of coverage can have the negative effect of giving the end user a false impression that they are searching something all-inclusive, when in fact they might be better off by searching a specialised database with superior indexing,’ he said.

A related challenge is that discovery services inevitably produce far more results that individual databases, which makes relevancy ranking and facets crucial. ‘The quality of relevancy ranking and ability to provide facets is linked directly to the quality of the metadata. If the end-user does a search and a million results are returned, how valuable is that result list if the most valuable results aren’t available on the first page? Without high-quality subject indexing, the user must sift through titanic result lists hoping to stumble on the most relevant articles, because they didn’t appear on the first page,’ said Brooks. ‘The one drawback of all discovery services, including EDS, is that there is no index browse for things like subjects or authors. However, EDS does provide facets for every search (such as subjects, authors, journals, source types and content providers),’ he continued.

User reactions

The work that discovery service providers are doing seems to be paying off. ‘We get comments back from users routinely through the feedback mechanism in Summon, and they tell us they love it. What’s more important are the results libraries are getting. A study at Grand Valley State University in Michigan, one of the first Summon users, shows that Summon is driving more usage of the library collection and opening users up to new content types. That’s the true test of whether we’re meeting our goal of getting end-users back to the library,’ said Gersch.

Ex Libris has had similar experiences with Primo: ‘Primo has been acquired by over 750 institutions worldwide, and the feedback we have received from libraries is extremely positive. Judging from usage statistics, one can clearly see that the number of searches in the library has gone up significantly while the average session time has decreased. At New York University, which reported an increase of 300 per cent in the number of searches after Primo was implemented, the average session length decreased from 15 minutes to six minutes. The fact that the sessions are shorter can be attributed to Primo’s relevance-ranking algorithm; according to the search logs at Seoul’s Yonsei University, titles that are required reading all appear on the first page of the result list today – whereas, before the implementation of Primo, they appeared as far away as the 15th page. At the same time, the number of searches grew by 400 per cent, testifying that the system has indeed been embraced.

Sam Brooks EBSCO

EBSCO’s Brooks said that the company has not experienced a product launch quite like that of EDS: ‘We have never had a product or service with such a high rate of conversion from trial status to customer status. I think that incredibly positive reaction of buyers (librarians) is directly related to the highly positive reaction of end users. Usability testing with end users has shown an amazing level of satisfaction. We have done and continue to do extensive end-user testing, often including heat maps that allow us to see where end users’ eyes are focused on various screens performing various functions,’ he said.

Plans and predictions

With such a positive reaction from libraries today, discovery service providers are excited about the future: ‘EDS customers have helped us decide on a development roadmap that includes: enhanced features when viewing catalogue records, consolidated book records, consolidated journal article records, etc. We are also including searchable cited references in EDS, which should be very exciting to faculty and researchers who publish papers. And by the end of 2011, EDS customers who also have Web of Science will have access to cited references for 16,470 unduplicated journals. We see EDS as a crucial part of what we do, and we will continue to invest in its features to better serve our customers,’ said Brooks.

Ex Libris has high hopes too: ‘I believe that the most important thing is to address the user’s context – institutional and professional affiliations, research area, past information-seeking behaviour, social network, and more. Systems that leverage such information along with intelligence gathered through usage metrics can help libraries provide better service to researchers. Knowing the user’s context and knowing the data – two areas that libraries excel in – will give libraries a huge advantage over web search engines and other sources of information,’ said Tamar Sadeh.

‘There is no doubt that scholarly information-seeking behaviour is changing along with people’s general information-seeking behaviour. Technological advances enable automated systems to be more context sensitive and to tailor services to the individual, thus blurring the distinction between human-human interaction and human-machine interaction. Specifically regarding information seeking, what systems “choose”, out of the huge information landscape, to put at the top of result lists is becoming a matter of focus and decision. I believe that the community’s real challenge is to leverage the developments that are likely to govern general information-seeking practices while retaining its scholarly focus.’

‘Users have no patience for information silos and it’s up to content creators, aggregators and publishers to align in a way that will better anticipate user needs and define paths to the best, most relevant content,’ concluded Michael Gersch of Serials Solutions. ‘I think we’ll see continuing elimination of barriers between content sources, turning the information landscape from reactive to proactive.’