CSMeD: Citation Screening Data for Cochrane Systematic Reviews

CSMeD is a collection of datasets for automated citation screening in systematic literature reviews, introduced by Wojciech Kusa at NeurIPS 2023. It provides labeled bibliographic data derived from Cochrane reviews to support the development of NLP models for medical research synthesis. The collection focuses on bridging the gap between manual screening processes and machine learning automation.

Use Cases

• Binary classification of study abstracts to automate inclusion/exclusion decisions
• Benchmarking NLP models on medical citation screening tasks
• Evaluating active learning algorithms to reduce manual screening workload in systematic reviews

Strengths

• Peer-reviewed at NeurIPS 2023
• Derived from Cochrane systematic reviews, widely considered the gold standard for medical evidence
• Released under the permissive Apache-2.0 license

Limitations

• Domain-specific focus on medical literature which may not generalize to other scientific fields
• Likely contains significant class imbalance as most screened citations are typically excluded in systematic reviews

Provenance

Source

Wojciech Kusa, NeurIPS 2023

Collection Method

Curated from Cochrane systematic review databases

Freshness

Last updated July 2024; based on a 2023 NeurIPS publication.

Geography

Global