Crowdsourcing and machine learning: finding lost trials together

Article type
Year
Authors
Noel-Storr A1, Thomas J2, Dooley G3, Mavergames C4, McDonald S5
1Cochrane Dementia and Cognitive Improvement Group, Oxford University
2EPPI-Centre, UCL. 'Transform' project; CQIMG
3Metaxis Ltd
4Cochrane Central Executive Team
5Cochrane Australia
Abstract
Objectives:
1. To outline the technical and methodological features of a pipeline that identifies research for systematic reviews – as soon as it is published.
2. For participants to have the opportunity to try some of the technologies described.
3. To promote discussion about the methodological and procedural issues this work raises.

Description: It is currently very time-consuming to find relevant studies for inclusion in systematic reviews. With ever-increasing volumes of research published each year, the problem of the data deluge is increasing. This workshop will open with a short introduction to two of the technologies developed as part of the ‘Transform’ project: i.e. the ‘evidence pipeline’ and ‘Cochrane Crowd’ platforms. Centralised searches will be conducted, and new citations fed into a machine learning platform. Automatic categorisation will determine the probability that a given citation is a randomized controlled trial and to which review/review group it is relevant. The Cochrane Crowd will then confirm/disconfirm the machine judgements creating a globally unique register of trials.

After a short introduction to the above technologies and processes, this workshop will facilitate the guided engagement of participants in trialling the online platforms. It will use this experience to kick-start a facilitated discussion regarding the methodological and organisational issues the new technologies raise. This is a Project Transform Team workshop.