'Screen For Me': harnessing the efficiencies of machine learning and Cochrane Crowd to identify randomized trials for Cochrane Reviews

Tags: Workshop
Noel-Storr A1, Thomas J2, McDonald S3, Dooley G4
1Oxford University, Cochrane Dementia and Cognitive Improvement Group, 2EPPI-Centre, UCL, London, 3Cochrane Australia, 4Metaxis Ltd


Machine learning and citizen science initiatives within Cochrane are transforming Cochrane's centralised efforts to identify reports of trials. The RCT Classifier and Cochrane Crowd are highly accurate at distinguishing between records that are likely to describe RCTs and those that are not. The 'Screen For Me' service enables individual Cochrane Reviews to take advantage of the efficiencies that this combination of human and machine effort offers.


Participants will:

- know when 'Screen For Me' is appropriate to use;

- gain a good understanding of the 'Screen For Me' workflow;

- try out the services on a training dataset;

- discuss the strengths and limitations of these new services.


This workshop is primarily aimed at Cochrane Information Specialists and Cochrane review authors. It will begin with a brief presentation of the 'Screen For Me' workflow and the pilot work conducted to date. The main part of the workshop will be hands-on. Participants working individually or in small groups will be supplied with a practice dataset representing the search results for a Cochrane Review. They will have the opportunity to send these results through the 'Screen For Me' workflow in CRS Web. Within the workflow, they will see the output from the RCT Classifier and also set up a task for Cochrane Crowd. The final 20 minutes of the workshop will be a question and answer discussion session.