A Simple Machine Learning Framework for Citation Screening of Aging and Longevity Research Studies
Description
This report describes the initial results of a research project to develop a machine learning framework to semi-automate citation screening in systematic reviews and meta-analyses. It was developed and evaluated in context of aging and longevity research studies. We conducted experiments on dataset related to a specific risk-benefit analysis. Empirical results show that by using the proposed system, reviewers save an average of around 60% of screening work compared to unaided screening to identify 95% of relevant papers. The logistic regression model used in our framework is one of the simplest machine learning models and very efficient to train. The resulting model is used for the update of the analysis, thereby offering an opportunity to reduce the reviewers workload. In addition, we use the results from cross-validation procedure to help identify potential screening errors made. We outline the framework and how it can be used in similar screening practices. The software is open sourced with dataset freely available.
Notes
Files
zenodo.4603365.pdf
Additional details
Related works
- Is compiled by
- Software: https://github.com/markolalovic/longevity-research-screening (URL)
- Is supplemented by
- Dataset: 10.5281/zenodo.4593916 (DOI)
- Presentation: 10.5281/zenodo.4594866 (DOI)
References
- A. Bannach-Brown, P. Przybyła, J. Thomas et al. (2019). Machine learning algorithms for systematic review: reducing workload in a preclinical review of animal studies and reducing human screening error. In Syst Rev 8:23.
- B.E. Howard, J. Phillips, K. Miller et al. (2016). SWIFT-Review: a text-mining workbench for systematic review. In Syst Rev 5:87.
- A. O'Mara-Eves, J. Thomas, J. McNaught et al. (2015). Using text mining for study identification in systematic reviews: a systematic review of current approaches. In Syst Rev 4:5.
- P. Przybyła, A.J. Brockmeier, G. Kontonatsios et al. (2018). Prioritising references for systematic reviews with RobotAnalyst: A user study. In Res Synth Methods, 9(3):470-488.
- B.C. Wallace, T.A. Trikalinos, J. Lau et al. (2010). Semi-automated screening of biomedical citations for systematic reviews. In BMC Bioinformatics 11:55.