Version 184.108.40.206 is a "minor" update on the surface, but beneath the hood, it actually hides major changes. In particular, with this release all the "Update review" features, which were driven by the large Microsoft Academic Graph database (MAG) are now tapping in a new data source: OpenAlex. This required to change vast parts of the EPPI-Reviewer "back end", which is invisible to users, but does in fact perform a lot of "work". The switch to OpenAlex was made necessary by Microsoft's decision to stop updating MAG, and made urgent by the number of surveillance projects that currently rely on the "Update Review" features.
Note: the "Update Review" features of EPPI-Reviewer are open to all EPPI-Reviewer users, but we do ask users to get in touch with us to make them available on a per-review basis, for now.
Apart from this major (and mostly invisible) change, version 220.127.116.11 includes the usual handful of new features and bugfixes.
New Feature (both versions): COVID-19 machine learning models
As part of our surveillance work for the department of health, we have been maintaining a regularly-updated evidence map of COVID-19 research. Subsequently, we started maintaining another "spin-off" map, focussed on research about Long-COVID. In turn, the work done on both maps allowed to collect sufficient data to train two new classifiers: the "COVID-19 map categories" which will assign references to the appropriate categories as seen in the comprehensive COVID-19 map, and the "Long COVID binary model", used to identify references about long COVID. Both models are now available to all EPPI-Reviewer users, in the same list as the other pre-trained, public machine learning models.
Please note: these models use a considerable amount of processing power, which means: each time they run, our "computing" bill goes up (considerably, if applied to many items) and they take a long time to run. Please apply them only when necessary, and be patient while waiting for the results. Submitting the same job multiple times will waste money, carbon and time, making things slower for no benefit.
New Feature: EPPI-Visualiser logs of activity
In EPPI-Reviewer Web, the "Setup Visualisation" page now provides access to the activity logs for EPPI-Visualiser, allowing review admins to find out how much the visualisations are visited, as well as monitoring how the data is being explored in a fairly granular level of detail. Of course, because of privacy concerns, the data collected and shown by this feature does not include any information about who is visiting visualisations.
New Feature: export references for offline screening on Title and Abstract
Upon request from Healthcare Improvement Scotland, we added a new "Export" option in the References tab (sub-menu of the "Export to RIS" button) in EPPI-Reviewer Web, called "HIS (ext. scr.)". This routine will export the currently selected references into an Excel file, reporting title, abstract, DOI, URL and multiple other (Journal-specific) fields. The last three columns provide space for external people to signal whether they recommend a given reference to be "included" or "excluded".
Please note that feeding back these decisions in EPPI-Reviewer is currently a manual process and we do not have immediate plans to automate it.
BugFix, Account manager: preventing unwanted "repeated actions"
As usage figures grow, and the number of users who have consistently used EPPI-Reviewer for many years keeps growing, a new problem with the account manager emerged. In some situations (invite people in a review, activate credit, and more) the Account Manager application might now take a little bit of time (seconds) to conclude a given operation, during this time it was possible to "click" on a number of "action" links, thus repeating the same command twice in a row, or issuing multiple commands while the data underneath was still changing. This could produce a number of undesirable side effects. To resolve this, we have updated the account manager so that it "disables" such "action links" while a given action is still happening.
BugFix, EPPI-Reviewer Web: run public classifiers
In version 18.104.22.168, public classifiers stopped "working" for EPPI-Reviewer Web; as soon as this bug was discovered, the problem was resolved with an "ad-hoc" patch that could be deployed immediately. With the current version, this problem is resolved with a definitive (better designed) solution.
BugFix, EPPI-Reviewer Web: apply classifier button
The button for "applying a classifier" in EPPI-Reviewer Web could on occasion be "Enabled" even if no classifier was selected, this problem has now been fixed.
Update: v. 22.214.171.124
This "flash update" resolved a backend problem that could prevent the application of machine learning models to the "Auto update" batches of items retreived from OpenAlex.