Discovery at Stanford Today

The Stanford University Libraries offer an enriched search experience dependent on multiple approaches and technologies that are based on constant contact with scholars and students. As Calter remarked in her SLA presentation, the content engineering behind the search interface looks to “go beyond the limitations of MARC records and offer deep linking to other content and media types.” Users can search the catalog, online articles, Yewno, and the library website, either separately or in one search.

Each tool has been purpose-built to respond to the input and direction provided by library users. These customizations are extensive and have involved the work of library staff over a period of years.

Calter reviewed some of the individual tools that the libraries use to enable richer search and enhanced record retrieval. Stanford was an early adapter of Blacklight, which Wikipedia describes as an open source “engine for creating search interfaces on top of [existing] indices. The software is frequently used by libraries to create discovery layers or institutional repositories.”

According to Hyeran Kang Brummett, a librarian and user experience expert from Indiana University, Bloomington, “Blacklight includes support [for] faceted browsing, relevance-based searching, bookmarking documents, [and] permanent URLs for documents” (blogs.libraries.indiana.edu/redux/2011/09/23/iucat-blacklight).

The enrichments enabled by Blacklight are integrated into the SearchWorks catalog. The catalog searches the physical collections and digital resources of Stanford’s libraries and includes books, media, and topic-specific databases. A customized version of EBSCO Discovery Service, called SearchWorks articles+, is also available and includes more than 300 million articles in hundreds of databases.

Early in the process, the library teams heard that users wanted data visualization tools. The Yewno ability to create a “knowledge map” was complementary to the other enabled search tools. In addition to the visualization tools, Yewno enables concept and topic exploration, automatic discovery and clustering, semantic indexing of full text, and multilingual discovery.

The library website enables users to “Find topic specialists, blog posts, descriptions of our notable collections, as well as libraries, hours, and policies.” If a user chooses to search all four tools, the results are presented in what Calter called “a Bento Box,” with a summary of results for each tool in a separate quadrant.

Records retrieved from the catalog show links to additional author/creator/contributor data, subjects, and supplemental links to items such as biographical information, publisher details, and tables of contents. For books, users can view items adjacent on the “virtual shelf,” emulating what many library users would do when taking items off the real shelf.

A Work in Progress

AI applications, including discovery, are developing rapidly, even daily. The technologies are improving, and users are being more specific about what they like—and don’t—regarding the new tools. Thus, local efforts to improve research tools are ongoing and creative collaborations are explored to enhance the efforts.

For example, Stanford participates in the Share-VDE Project (Share Virtual Discovery Environment). Share-VDE is a cooperative project of 17 leading academic institutions based on the Casalini Libri and @Cult partnership. The project is described as “a prototype of a virtual discovery environment with a three BIBFRAME layer architecture (Person/Work, Instance, Item) [that] has been established through the individual processes of analysis, entity identification and reconciliation, conversion and publication of data from MARC21 to RDF, within the context of libraries with different systems, habits and cataloguing traditions” (share-vde.org).

Calter looks forward to Share-VDE enabling the participating libraries to link and automate reconciliation of vocabularies and transform the collective metadata generation and visibility. According to the Share-VDE website, the extraction of persons and works is “enriched with data from external sources for extending the research potential” while “the Instance level aggregates the resources of the various editions, even in different languages, and connects them with the overlying Person/Work.” The Share-VDE processing completes the loop on the Item Level by connecting “each Instance (Publication)... to the information contained in the dataset of the local OPAC of each single [participating] library” (share-vde.org/sharevde/clusters?l=en).

At Stanford, there is an ongoing best practice dialogue about AI. The Library AI Conversations series (library.stanford.edu/projects/artificial-intelligence/library-ai-conversations) brings AI researchers into the library to present their work and initiate discussions. These are monthly meetings featuring a guest speaker, with the objective “to increase our shared understanding of how machine intelligence can serve the library.” Conversations such as these are critical to the library being aware of users’ needs and preferences as well as technical developments to enhance discovery functionality. Different paths are explored and tested. Perhaps a hybrid will emerge to best serve the research community. In fact, Calter said, she “does not see one search tool encompassing all approaches.”

As I left the meeting, I came away feeling encouraged about what the Stanford University Libraries are doing to create a powerful differentiation between library content and services and the open web-dominated information environment.

They are leaving no stone unturned—better communications and collaborations with constituencies and vendors, enhancing and expanding what can be revealed by metadata, careful tuning of user interfaces, and turning to emerging technologies such as AI and graphic display of research results. These are all good things that reinforce that the library has the good stuff, and it is certainly worth the time to use and learn tools that take your research to the next level.