Libraries must accommodate a variety of expectations from the patrons who visit their physical facilities or virtual offerings. One dimension of this experience relates to satisfying those with something specific in mind. These patrons may have a favorite author or topic in mind and want specific resources, which might include the next book by that author or an exhaustive set of materials in the bibliography of a research topic. The catalogs and discovery services that libraries present to their users are designed especially for these kinds of information fulfillment activities.
Sometimes, users come to the library without a specific item or topic in mind, but want to peruse the shelves or displays in hopes of coming across something interesting. How many times do we visit a library or bookstore expecting that something interesting will catch our attention? Libraries can facilitate this kind of serendipitous exploration in the ways that they organize their physical collections and design user spaces. For example, many public libraries have adopted some of the ways retail establishments display books and media. In addition to the standard shelving with only the book spines exposed on the shelf, featured items are displayed with the front of the book facing out and placed in prominent spaces in the library.
Some libraries set aside a Lucky Day Collection in which some of the most popular items that tend to be constantly checked out and have long hold queues are made available for those who come to the library in person. These types of techniques help draw community members into the library and often delight them with reading materials they would not have planned to look for in advance.
Replicating serendipitous or fortuitous discoveries in a library's virtual offerings isn't easily accomplished. Library catalogs, discovery services, and websites tend to do better at guiding online visitors to items representing titles, authors, or topics that match what they have entered into a search box. But it's important for library interfaces to go beyond mechanical search and retrieval and help patrons explore library collections in more creative ways with at least some of the qualities of what is possible when they visit the physical library.
A search box functions as one of the starting points on the library's website or virtual branch. Visitors are used to search-it's a basic feature of almost all web destinations and social networks. Libraries, along with the vendors that develop library software, have been struggling since the advent of the web to create effective search technologies that live up to the expectations set by the commercial sphere. The evolution from online catalogs to federated search tools to the current phase of index-based discovery services has produced important progress in creating faster, more comprehensive, and more effective environments for helping patrons explore library collections, although much work remains in perfecting these products.
In an ideal world, a search tool connects users with the very best items that match the intentions behind their query. Many commercial sites not only excel at forgiving typos and misspellings, but also thoroughly exploit information about the user and other clues of context in order to present satisfying search results. While library-oriented systems may not have the seemingly unlimited financial and engineering resources available in the ecommerce sector, we must nonetheless hone technologies to enable library users to not only search collections of reliable resources and receive objective results, but to also provide ways to explore these collections through many possible tangents that veer off in unpredictable directions.
Relevancy attempts to return items in response to a search that qualify as result candidates and also are presented in an order in which the top item most closely adheres to what the patron had in mind. The best relevancy may not necessarily follow a strict or mechanical algorithm, but it may include an almost predictive quality. What a user types into a search box may not necessarily correspond to the query that would elicit the best result set for the intended question or topic. Apart from misspellings or typos, users cannot be expected to know the ideal terms or text that match the subject headings, vocabularies, or other controlled terms built into library metadata.
I observe several threads of development in the current slate of index-based discovery services. One area involves expanding the content used to populate the indexes to include more publishers and to close any remaining gaps in unrepresented scholarly content. It's certainly a problem when resources to which a library subscribes cannot be found through its discovery service. Discovery service developers also constantly work to refine the relevancy algorithms to prioritize and highlight the most important and interesting items in a result set. Other aspects of development focus on tools and techniques outside of the realm of pure search and retrieval, such as recommendation services, improved methods for linking or presenting full text, assembling resource pages, and other interface improvements.
It is natural for in-person library visitors to browse the shelves to discover items of interest. The fundamentals of organizing library collections include assigning classification numbers so related materials will appear on the shelf together and authors and topics will be ordered in optimal ways for patrons to see all the items available in a given area of interest. The effectiveness of in-library browsing is limited, since patrons will not see items that are checked out, that happen to be placed in other branches or sections, or that have been purchased but not completely processed. Such items would be available for request, although not physically present on the shelf. As libraries move increasing portions of their research collections with less immediate demand to off-site storage, what's on the shelves becomes even more of an underrepresentation of a library's collection.
Online representations of a library's collection have the potential to provide an even better experience than the physical facilities. Catalogs or discovery systems can virtually assemble items in shelf order, consolidating them regardless of circulation status and across all shelving and storage locations. Offering lists of items within a given classification range or shelf order has become a fairly routine feature of library catalog interfaces. Providing this functionality visually with cover art images can be especially effective for public libraries. Modeling the experience of visiting the library in person, a user might begin by searching for an item-perhaps a book he or she read previously-and then touching some control to browse through virtually adjacent items. Physical and virtual browsing depends on careful attention to the classification of materials and assignment of call numbers.
Book carousels present selections of collection items, especially books and media, using cover art rotating visually in a way that invites visitors to the website to select items that might seem interesting. The items presented in a book carousel can be selected in various ways. Some may be produced out of lists from different categories of best-selling or award-winning titles. It's also possible to populate a visual book carousel with titles selected by local librarians or by those identified by circulation reports as having frequent checkouts. Libraries can promote items by local authors, on local topics, or on history that may appeal to a specific library's clientele. When selecting items to promote in this way, it's essential to present those that are actually available. It can be frustrating to a patron to select a featured item and discover that it is not owned by the library or has an impossibly long hold queue.
Many patrons may ask a librarian for advice on other titles that might be available on a topic of interest. Librarians can tap their personal knowledge on a topic and what a patron says about the nature of the topic to suggest other items owned by the library or that may be requested. In the virtual world, we experience mechanically generated recommendations on sites such as Amazon and most other ecommerce sites based on current selections, previous purchases, and data mined from often unexplained aspects of our browsing history across related and unrelated sites. The patterns seen by the general population of users can also inform recommendations. When placing an item in an online shopping basket, we often see, "Others who bought this item also bought these items. ..."
Recommendation services have become an expected feature in library catalogs and discovery services as well. Such recommendations can be generated mechanically based on other items that are by the same authors, that have been assigned similar subject headings, or that have some other element of metadata in common.
It's also possible to generate recommendations based on the selection and circulation patterns collected across all users of the system. These use patterns may reveal related materials beyond what would be apparent in descriptive metadata. If an expert user selects multiple items during an online session, it can be inferred that the items might be related even if they don't have exactly the same metadata terms. Stronger inferences can be made when the topic or discipline can be layered into the recommendation algorithm. For example, if you are an organic chemist, use patterns previously seen for others in the same field would be especially valuable.
Any data collected about user patterns can be valuable in helping a discovery service improve the presentation of results by relevancy and in making recommendations. At the same time, these systems are expected to treat personal data carefully. Recording patterns of selection by categories of users may not cross the lines of privacy concern compared to those that preserve individual online behavior. One of the challenges in developing library services that are able to provide personalized services lies in doing so based on patterns seen in categories of users established in ways that do not intrude on the privacy of others who use the system.
Recommendations from other patrons with similar interests can be especially helpful. In the same way that individuals may depend on suggestions from family members, friends, and colleagues, a virtual-library environment might be able to connect patrons in networks of similar interests and enable them to make recommendations to each other. This kind of social flavor is often seen in web destinations. Comments, ratings, and rankings given by other users provide helpful information in choosing movies, restaurants, travel accommodations, or almost any other service. In a library context, such user-to-user interactions can likewise be helpful, although they need to be provided through a more carefully defined privacy framework. Patrons may select profile options that enable them to reveal their personal preferences or identity or they may be able to use these features anonymously.
Offering Plausible Tangents
The techniques mentioned so far are examples that libraries include in their virtual presence to help patrons explore their collections. When beginning with a search, it's essential to provide complete and relevant results, as well as to lead to materials related in ways not entirely apparent through strict criteria of metadata-but revealed through social use patterns. Others may want to explore the collection the same way they do the physical library: Find a familiar starting point and just look around for interesting items. One might select a title, read the blurb on the jacket, and move on if it isn't especially interesting.
It's incredibly challenging to design a virtual-library environment that includes at least some kind of serendipitous discovery. Search algorithms are continually being refined to improve the scope and accuracy of search, but this can't be the only aspect of collection discovery. One can offer prepopulated book carousels or mechanical recommendation systems that may present useful items to the general user population. It is exceptionally difficult to design systems that go beyond these more rigid approaches and offer a softer approach that enables exploration in less predictable directions. Such exploration isn't based as much on mechanical relationships among content items, but more in the ability to present plausible tangents. These tangents can't be defined in advance for all users, but are offered according to individualized circumstances based on a variety of subtle clues.
The next phase of library interfaces will benefit from increased reliance on artificial intelligence and better understanding of the psychology of user behavior. Those with deep expertise in the art of user experience are well-positioned to help shape library environments and virtual branches to extend beyond mechanical approaches and deliver with a personal and human touch.