At Penn State, we’re planning a discovery revamp which will include presenting our catalog with a Blacklight layer over the MARC records and creating a Bento landing page for the front-page search, which currently goes to Summon. We’ve been reviewing other institutions who’ve implemented such approaches, determining their choices, and assessing how they do or don’t meet our own needs.

The box causing the most conversation right now has been our Summon search. We don’t plan to serve it up in Blacklight yet, as some institutions do, although that’s on the table for the future. But, if our Summon currently contains 1,000,324,214 records1 (whereas our library’s catalog contains almost 7 million…) what should we actually be searching when we’re returning those results from Summon to the Bento page.

Of the institutions we reviewed, all had a box labeled something along the lines of Articles or Articles+ (or Articles & Ebooks). Most did not search the entire serials/databases/subscription interface but used some kind of faceting to filter result types toward a more scholarly result. Some excluded. Some included. Some did both. I found the scan of enough interest that I decided to share them on both our staff blog and my personal website.

I’m including two tables from the scan. The first is a table of information gathered about choices for the Articles/Articles+ Bento box. The second is a table which compares the total possible number of results from an institution with the total number of results when faceting is applied. The second table only includes institutions which provide at least fairly direct access to their Summon. I couldn’t do a null search of Summon over Blacklight.

This iteration of the scan was conducted between 16 and 17 January, 2018 and numbers, choices, and percentages may change.

Articles+ Scan

These results were all collected as an off-campus user who was not logged in. Institutions included in the scan were those whose Bento implementations we had chosen for review, not all institutions who may be using such functions. Ordering has no meaning.

Institution Equivalent Box Faceting Product/Notes
Princeton Articles+ INCLUDE: Scholarly & Peer Review Summon
Dartmouth Articles INCLUDE: Scholarly & Peer Review, EXCLUDE Book Review; EXCLUDE: Newspaper Article Summon
Stanford Articles+ (subtitle: Journal articles, e-books, & other e-resources) INCLUDE: Direct access to full text Summon (Blacklight)
Cornell Articles & Full Text INCLUDE: Items Available to Cornell, EXCLUDE Newspaper Articles Summon, behind a shibboleth.
Yale Articles+ (i) 1″ Contains articles, eBooks, electronic resources” Yale’s Collection Only. URL includes the term “articles,” but it appears to just be a Summon skin without filtering out of books or ebooks.
Brown Journal Articles (i) “Full text articles from scholarly journals, conference proceedings, book chapters, and other electronic content from the library’s subscription resources” Made local inquiry. Search returns facets for Archives/Manuscripts, books, Computer File, Journal Articles, Map, Musical Score, Periodical Title, sound Recording, Thesis/Dissertation, Video, Visual Material. Null search of front page not easily possible. EBSCOHost, behind a shibboleth.
Duke Articles (subtitle: from journals and magazines) INCLUDE: Journal Article, INCLUDE Magazine Article Summon
Smith Articles INCLUDE Academic Journals EBSCOHost
Columbia Articles (i) “Articles, e-books, dissertations, music, images, and more from a mostly full-text database” INCLUDE: Columbia’s Collection Only, EXCLUDE Newspaper Articles Summon
UI-UC Articles (i) “EBSCO Databases included in this search: [see next section]” Databases: Academic Search Complete; Business Source Complete; Library, Information Science & Technology Abstracts; Communications & Mass Media Complete; America: History & Life; Historical Abstracts; SPORTDiscus; ERIC; SocINDEX with Full Text EBSCOHost, behind shibboleth.
Penn Articles+ No faceting or filtering Summon
Michigan State University Articles (i) “From journals, magazines & conferences” INLUDE: Journal Article, Magazine Article, Conference Proceeding Summon

Edit, 2018-01-25: Added Michigan State University

How Faceting Affects Search Results

These results were all collected as an off-campus user who was not logged in. Institutions included in the scan were those whose Bento implementations we had chosen for review which also had Summon and had not applied a Blacklight layer. Ordering has no meaning.

Institution Total possible With Filter Filter Type Percentage
Princeton 998,667,021 81,633,146 INCLUDE Scholarly & Peer Review 8%
Duke 1,000,746,600 183,955,317 INCLUDE Journal Article, INCLUDE Magazine Article 18%
Yale2 817,076,655 817,076,655 Yale’s Collection Only. 100%
Dartmouth 878,541,266 61,899,637 INCLUDE Scholarly & Peer Review, EXCLUDE Book Review, EXCLUDE Newspaper Article 7%
Columbia2 1,051,600,796 421,991,628 INCLUDE, Columbia’s Collection Only, EXCLUDE Newspaper Articles 40%
Penn 825,589,737 825,589,737 n/a 100%
Michigan State University 962,179,590 151,112,768 INLUDE: Journal Article, Magazine Article, Conference Proceeding 15%

Edit, 2018-01-25: Added Michigan State University

Our Own Numbers

We did our own experimentation on the subject and determined that faceting our Summon to include only Scholarly & Peer Review material returns 97 million results. Of those, 9.8 million are Book Reviews, which are a known issue in clouding our search results (although some Book Reviews are tagged as Journal Articles in Summon). So, cutting those out, we come down to almost 88 million records. We’ll be doing user testing, of course, so these decisions aren’t solid. But I think they’re a good starting place for us.

Notes

1. Results of a null search and a topic for another day. This is also not unusual. Update: Aaron Tay shared with me the numbers he got when he did a similar scan back in 2012 and you’ll see it’s an enormous increase in sheer quantity in 5-6 years.return

2. While Yale and Columbia’s facet limit to “[Institution’s] collection only,” several other institutions have opt-in of including results from other institutions (vs. this opt-out). Although this exploration involved clearing all checkboxes, it was felt that leaving this box checked effectively compared institutional collection numbers. Also, it should be noted Yale and Columbia then made different choices about other facets, changing their %s. return

2 Comments

  1. I must say I initally didn’t quite understand what you are trying to do here with the %.

    In a bento, I would think the focus should be on the number and types of content boxes you should include.

    The largest Summon instances will be around 800+ to 1000+ million items and most of them will be from the shared Summon Central index which will drown out even the largest library’s catalogue. But that’s why you have a bento box right? To solve that issue.

    But I think you are implying that these Universities by choosing to filter Summon output in their bento (and not surfacing the other content in other boxes) are implictly saying the rest of the Summon index is of low value to them so it doesn’t appear.

    Frankly, I think the peer reviewed portion of the article is where most of the value of the index. Newspapers to some extent for some disciplines. Everything else is mostly trash, particularly book reviews (and as you pointed out these are often misclasified as artiles.

    Reply
    • Yes, I was attempting to show what % of their entire Summon index people are actually searching/judging to be the most useful. One thing we did with this data was experiment by seeing what our search results looked like after we filtered on various criteria. For example, 9 million book reviews were too many (given our negative user feedback about them). We knew that Newspapers can also cause an issue because of our large newspaper holdings (239 million), however once we filtered to Scholarly & Peer Review (the name of our facet), we found so few newspapers were included that we left them. However we may decide to filter them out if it comes up in user testing. We’ll see!

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *