goo.gl/74nU1b A Decade in Discovery Updating Assumptions and Conclusions in Information Access and Retrieval A Decade in Discovery Bill Mischo (Illinois) Mary Schlembach (Illinois) Michael Norman (Illinois) Bruce Barton (Wisconsin) Lee Konrad (Wisconsin) Tony Zanders (EBSCO) goo.gl/74nU1b

“We bring the interface. You do the rest.” Discovery in 2007 Division of labor (libraries) “We bring the interface. You do the rest.” Discovery in 2007 Improving algorithm Publisher relations Metadata enhancement Linking issues Interoperability issues Third-party integrations Post-implementation support Custom apps

“You bring the interface. We do the rest.” Discovery in 2017 Division of labor (libraries) Selecting/designing interface Localizing the experience

EDS API Features Did You Mean Autocorrected Search Research Starters Direct Links to PDF EDS API Features

EDS API Features Did You Mean Autocorrected Search Research Starters Direct Links to PDF Publication Exact Match Enhanced Subject Precision Citation Search EDS API Features

Peters BS, Jaoko W, Vardas E, Panayotakopoulos G, Fast P, Schmidt C, Gilmour J, Bogoshi M, Omosa-Manyonyi G, Dally L, et al. Studies of a prophylactic HIV-1 vaccine candidate based on modified vaccinia virus Ankara (MVA) with and without DNA priming: effects of dosage and route on safety and immunogenicity. Vaccine. 2007;25:2120–7. doi: 10.1016/j.vaccine.2006.11.016.

EDS API Features Did You Mean Autocorrected Search Research Starters Publication Exact Match Enhanced Subject Precision Citation Search Article Image Preview Citation export EDS API Features

EDS API Examples Bento @ MIT libraries.mit.edu Bento @ Smith smith.edu/libraries Bento @ U Alabama lib.ua.edu Bento @ U Illinois library.illinois.edu Blacklight @ Stanford searchworks.stanford.edu/articles Blacklight @ U Virginia library.virginia.edu Stacks @ CalTech library.caltech.edu Drupal @ Rice library.rice.edu Custom UX @ BYU lib.byu.edu EDS API Examples

Innovations in Discovery Systems: User Studies and the Bento Approach William H. Mischo Michael A. Norman Mary C. Schlembach University of Illinois at Urbana-Champaign Library Grainger Engineering Library Information Center The Charleston Conference November 8, 2017

Library Discovery at a Crossroads Academic libraries should “step back to reconfirm (or reconsider) their vision for discovery, to ensure that their visions connect with information-seeking practices and preferences, and to determine whether they have a viable strategy in place … to achieve their vision” Roger Schonfeld, Ithaka S+R, 2014

Full Library Discovery Coined by Lorcan Dempsey 2013 refers to features that move beyond the retrieval of collection materials to also include local and connecting information services and content Includes library websites, local LibGuide information, subject specialist links, and course management system content

Evolution of Discovery Over the last 30 years: Supercatalogs with A&I services loaded Federated search systems employing broadcast searching Web-Scale Discovery Systems (WSDS) with metadata and full-text content aggregated into a single consolidated index Hybrid bento style systems with information results presented in a zoned screen display with content grouped by type/material

Classic Web-Scale Discovery Recent studies on WSDS have identified usability issues: Confusion with blended results displays relevancy rankings burying known-items inadequate access to local library services Need to better address known-item searching Expediting access to full-text

Bento Advantages Partitions results by material type or format Addresses issues with WSDS blended results and relevancy ranking problems Lends itself to full library discovery by displaying vendor/WSDS API results and local information services and content Library websites, pathfinder information, local services, subject specialist links, CMS content One click links to full-text or to publisher presentation pages (bypass link resolver)

Discovery at Illinois Evolved from federated search through web-scale discovery to bento (hybrid aggregated and federated) Search assistance and suggestive prompts Custom transaction logs and user behavior studies

Illinois Transaction Log Analysis July 2016 – June 2017 (latest) 1.49 million searches, 1.15 million clickthroughs High number of words per query – 5.11 (up from 3.76 & 4.11) 10% are one word, 24.7% > 7 words, 49.2% were <= 3 words Large number of copy-and-paste of title, author/title, full citations DOI searches (73 searches per day) In sample of 4100 searches, Known-item searching is high = 56% Users often have a material type in mind when they search Use of search assistance high Limit operations using Gateway tabs to material types used = 24%

Example Searches from ES TL Competent Jerks, Lovable Fools, and the Formation of Social Networks The virtual supermarket: An innovative research tool to study consumer food purchasing Mothers and fathers : a study of the development and negotiation of parental behaviour Matouschek, Kellis, Serrano, Fersht Nature Kinsella and Phillips 2005 Hemenway, D. (2010). Why We don't Spend Enough on Public Health. New England Journal of Medicine, 362(18), 1657 university, need for money J. Quant. Spectrosc. Radiat. Transfer.

Illinois Bento features Employs context-specific and adaptive search assistance Spelling suggestion links, Libguide links, direct links for frequent searches, limit suggestions, DOI identification, online chat link, journal title links Direct link to PDF when available DOI, publisher, SFX links for articles, e-book links from catalog Altmetric badges Writes out custom transaction logs

Illinois Bento Box Discovery Model Keywords  {search terms} GO Search: Suggestions: Direct link to {search term} Look at the Journals or databases that match this search. DOI direct link Limit this search to title words. Need additional assistance? Ask a Librarian Classic Easy Search Articles (EBSCO EDS API) 10 Results See all matches EBSCO PDF links, DOI links, Publisher links, Catalog links, Scopus record Library Catalog (VuFind) 10 Results See all matches Direct links to full-txt Subject Top Picks Subject Database matches Departmental Library Staff contact General Searches Library Web Pages Reports, Collections, Pages Articles: Additional A&I Service API (Scopus, Engineering Village, IEEE) 5 Results See all matches DOI links Advertise Library Service Click image

Bento Clickthroughs by Category User Clickthroughs by Category Percentages Article links 56.1% OPACs and Books 33.6% Full-text clicks 16.6% Suggestions 9.2% Journal Title links 3.6% Library links or contacts 0.2%

Bento Characteristics 38 academic libraries with bento instances All have Books and Articles areas WSDS for articles: 15 Summon, 4 Primo, 10 EDS Other resources: Website search = 17 Library/research guides = 12 Journal titles = 12 Databases = 14 Digital collections = 11 Subject Top Picks = 10 Contacts = 10 OPAC often separate application: VuFind, Blacklight

Bento Libraries NCSU, Illinois, Michigan, Columbia, Duke, Stanford, Johns Hopkins, UNC, Virginia, Cornell, Princeton, Toronto, Vanderbilt, Yale, Penn, Brown, Dartmouth, Alabama, Rice, Cal Tech, Oklahoma, Smith College, Villanova, Wayne State, George Washington, Calgary, Alberta, Simon Fraser, North Texas, Boston College, Bryn Mawr

Bento Observations Have Local control and customization Features sets vary; many bento versions do not do spell checking and do not provide direct links to full-text Only three employ the one-click to full-text w/p going through local link resolver Have Local control and customization But, requires programming/server staff and maintenance … API processing Need for catalog item availability and direct links to ebooks

Coordinated Discovery University of Wisconsin-Madison

Motivations and goals

Discovery challenges University libraries support many types of resources. Different types of resources imply: Different modes of search Different modes of evaluating search results Different modes of delivery Respect the differences in the discovery experience! But this makes for disjoint, siloed, uncoordinated discovery.

What is coordinated discovery? Given separate resource categories optimized for the resource: 1. Provide good navigation between categories. 2. Carry forward queries from one category to the next. 3. Suggest relevant items from other categories based on the current search.

Suggestions Mini-bento box? Yes and no. Group by type bento-style. But filter by closeness of fit to beyond relevancy ranking. + Hand-curated suggestions.

Closing note We’re in the middle of implementing a common design and behavior vocabulary across our discovery platforms. The aim is to leverage our patrons evolving experience with discovery on the web in the library context.

goo.gl/74nU1b A Decade in Discovery Updating Assumptions and Conclusions in Information Access and Retrieval A Decade in Discovery Bill Mischo (Illinois) Mary Schlembach (Illinois) Michael Norman (Illinois) Bruce Barton (Wisconsin) Lee Konrad (Wisconsin) Tony Zanders (EBSCO) goo.gl/74nU1b