Classifies COVID-19 records for the Cochrane register
The Cochrane COVID-19 Study Classifier is machine learning tool to assist in the screening process of titles and abstracts of studies. It identifies key features and patterns in the text that indicate whether a study is likely to be relevant to COVID-19 research, based on a large dataset of previously classified studies. In the current workflow, records identified from PubMed and Embase are run through the classifier. This allows it to automatically categorize new publications and prioritize those that are most likely to be useful for inclusion in the Cochrane COVID-19 Study Register, reducing the need for manual screening. While the study classifier assists at title and abstract screening, the included records are sent for manual screening. The classifier underwent calibration to ensure it met a target recall rate of over 99%, with a precision rate of about 63.8%.
