Hi Phil,

EPPI Reviewer can use Machine Learning to assist with screening (referred to as "Priority Screening" in the existing version of EPPI Reviewer - ER 4), but we have yet to completely port this feature over to the new (beta) version of EPPI Reviewer - ER Web.

However, as you will be working on updates and already have a set of "training data", the "Classifier" function already available in ER Web should work for you.

To use your existing references as training data and then effect priority screening for an update -:

1. Upload your included studies from your previous review / training data, and assign them to an INCLUDED code in your screening codeset.
    
2. Upload your excluded studies from the previous review and assign to the EXCLUDED code.
   
   
3. Upload your new search results / the new updates to your reference lists (as yet unscreened).

Then you can start priority screening. The classifier will run and build a model based on your old review data, hopefully bringing all the relevant items from your new references to the beginning of the screening list...

I will email you directly with some screenshots of this process.

(Information about Machine Learning and Priority Screening within EPPI Reviewer can be found at https://eppi.ioe.ac.uk/CMS/Portals/35...al.pdf#page=115.)

Kind regards,

EPPI Support

Dear Alison,

The relevant section in the ER 4 manual is around page 115 (see https://eppi.ioe.ac.uk/CMS/Portals/35...al.pdf#page=115 for direct link).

We also have a separate document on Machine Learning in ER Web, available at https://eppi.ioe.ac.uk/CMS/Portals/35...web_version.pdf.

(With respect to selecting references and giving them all a particular code, in ER 4 you can select the relevant items, then right-click on a particular code and can assign that code to all your selected items (or any selected search results). (Select the "Assign selected items to this code" option.)

In ER Web, you can select the relevant items, then select your chosen code. Finally, click the Assign Code button at the top of the code tree.

I will reply more comprehensively via email. (You may find it easier and quicker to get support on specific queries via our support address EPPISupport@UCL.ac.uk, rather than the user forum.)


Best wishes,

Zak