I participated on a MLA 2021 panel on problems with full-text linking. My fellow panelists were Lindsay Barnett, Basia Delawska-Elliot, and Angela Spencer while Andrew Hickner moderated the panel as well as lead our preparations. Below are my slides, excerpted from our larger presentation, and some prepared text that accurately represents my remarks. I’m unaware of a publicly available recording of the session but, if I find one, I will post it here.
Slide 1: Research on Discovery Layer Linking
I will focus more on the technical aspects of full text linking.
In the Fall of 2019, I performed a research study where I tested hundreds of links from our discovery layer to see if they successfully navigated to the full text. Last Spring, I repeated this study after an underlying index change and obtained near identical results, increasing my confidence in the experimental design.
While I obtained interesting data on the success rates of database platforms and content types like book reviews versus reference sources, there were two primary takeaways for me. First, only about 80% of links succeeded, much lower than I had hoped. This proves our linking issues are not isolated complaints but systemic problems.
Second, OpenURL links fared far worse than Index Direct links. OpenURLs are links where a source sends article metadata to a link resolver which knows your library’s database subscriptions, trying to connect to full text. OpenURL links are common not only in discovery layers but also in A&I databases. In an index direct link, on the other hand, the discovery system itself knows something about the destination resource and thus bypasses the link resolver, going straight to the full text.
Slide 2: Patron-reported Broken Links at CCA (2018 - present)
But why do links break? There are many reasons because full text links rely on such disparate data, software, and institutions. Here is a breakdown of reported broken links at my institution. Let me explain our most frequent ones.
General linking / OpenURL problems are the most common. These are identifiable when modifying a linking strategy or the format of an OpenURL solves the issue. Secondly, often an article will not exist in the source database, whether due to vendor omission or because of knowledge base configuration problems like inaccurate subscription date ranges.
Third, we’ve found that specific platforms can pose problems by having faulty or limited linking setups. We have had problems with EBSCOhost and LexisNexis specifically. Finally, there are issues attributable to bad metadata, whether that metadata is in the linking source or the destination database.
Some other problems that I don’t have time to cover in detail are granularity mismatch, EZproxy configuration, title-level links, and journal embargoes.
Slide 3: Tangible Improvements to Full Text Linking
What do we do to address problems with full text access? I try to focus on identifying whole classes of problem as opposed to solving issues one-by-one, because fixing inaccurate metadata for every single article is not scalable. For instance, we have found entire journals with poor linking strategies or malformed publication dates. If I tell the vendor’s support team about such large issues, they can fix hundreds or thousands of links at a time.
Secondly, we want users to report broken links and to feel confident we can help. We have a reporting function I built into our discovery layer which is the source of the pie chart on the previous slide. But addressing broken links can also be done at the reference desk, where librarians teach website navigation and work around access obstacles. For my linking studies, I could eventually find the full text for approximately fifty percent of broken links. When machines fail us, human assistance is invaluable.
Finally, if certain resources are the source of repeated problems that the vendor is unable or unwilling to address, you can improve your user experience and the efficacy of your budget by dropping the subscription. We adjusted our subscriptions a couple times due to linking concerns with noticeable positive results. I realize some subscriptions are held holy but remember you may be receiving only a fraction of their promised value.
Could you talk about the role metadata quality plays with link resolvers? Is it the main factor or are there other things that could go wrong?
In general, metadata is but one of several potential issues with access to full text. At my institution and others, I’ve seen figures that metadata is at the root of somewhere between one-fith to one-third of all linking problems. That may be a large share but it’s not the majority. Futhermore, libraries don’t have control over most metadata problems. Sometimes there is a disagreement between two vendors where neither can be convinced to change their record(s), leading to an incorrigible problem.
Tell us how you use reports as a way of gathering data?
User-generated reports were the source of the pie chart data on my second slide. We use reports not as the final say on what our major problems are and where our focus should be but as a mere indicator of where to look. Reports sometimes help highlight larger problems but also can be one-off issues that are quickly resolved.