Using text mining for study identification in systematic reviews: a systematic review of current approaches. O’Mara-Eves A et al. Syst Rev. 2015 Jan 14;4:5
I was really pleased to see this paper as it is, itself, a systematic review. I was also pleased to see it comes from the Evidence for Policy and Practice Information and Coordinating (EPPI)-Centre – a really innovative centre of activity!
The aim of the paper states:
“The evidence base around the use of text mining for screening has not yet been pulled together systematically; this systematic review fills that research gap….the review aims to increase awareness of the potential of these technologies and promote further collaborative research between the computer science and systematic review communities.”
To put it another way, systematic reviews tend to find – via a very sensitive search – thousands of potential articles. These typically need to be screened via a manual process, typically involving two screeners, to find which articles actually need to be included in the review. It is a very time-consuming (and therefore costly) system. If we want to speed evidence synthesis up this is an area ripe for innovation and, as the paper shows, there is much work being undertaken with some very promising results.
The review is focussed on 5 key questions:
- what is the state of the evidence base?
- how has workload reduction been evaluated?
- what are the purposes of semi-automation and how effective are they?
- how have key contextual problems of applying text mining to the systematic review field been addressed?
- what challenges to implementation have emerged?
The paper is too large to summarise here (so hardly an ‘article review’, more an ‘article alert’), read via the URL above but they conclude:
“Whilst there is a relatively abundant and active evidence base evaluating the use of text mining for reducing workload in screening for systematic reviews, it is a diverse and complex literature. The vast array of different issues explored makes it difficult to draw any conclusions about the most effective approach. There are, however, key messages regarding the complexity of applying text mining to the systematic review context and the challenges that implementing such technologies in this area will encounter. Future research will particularly need to address: the issue of replication of evaluations; the suitability of the technologies for use across a range of subject-matter areas; and the usability and acceptability of using these technologies amongst systematic review (non-computer scientist) audiences.“