Automated Systems for Systematic Review Processes

The summarising describes the application of artificial intelligence and computational methods to streamline each stage of a systematic review: from literature retrieval and screening to data extraction and synthesis. Automated systems employ machine learning algorithms, natural language processing and active learning to prioritise relevant records, reduce manual workload and enhance reproducibility. Tools range from web-based platforms that assist with title and abstract screening through predictive modelling to pipelines integrating large language models for rapid summarisation and evidence mapping. Adoption of these approaches has accelerated the production of living evidence syntheses, facilitating continuous updates and reducing research waste. Integration with collaborative software has enabled transparent decision-making and global access, supporting evidence-based policy and clinical practice across disciplines.

Research from Nature Portfolio

A recent development has introduced an open source active-learning pipeline designed to assist with title and abstract screening in systematic reviews. By iteratively updating the training set based on user feedback, the system rapidly converges on relevant studies while maintaining high recall. Simulation studies demonstrated substantial reductions in screening effort compared to manual review, and user evaluations confirmed improvements in efficiency and transparency. The software’s modular design allows extension to various domains and invites community contributions to refine algorithms and workflows.

Research from all publishers

One study applied zero-shot classification methods to automate abstract screening, enabling the system to assign inclusion labels without task-specific training data. The approach achieved competitive precision and recall across multiple public datasets, indicating potential to reduce screening burdens and human error. In parallel, evaluations of web-based screening tools have compared user experience and feature sets across platforms. Two leading applications consistently emerged for title and abstract management: they offer collaborative workflows, bulk annotation, conflict resolution and real-time analytics, yielding time savings of over 40 % and fostering team-based review processes.

Automated Systems for Systematic Review Processes publication trend

The graph below shows the total number of articles in automated systems for systematic review processes across all publications each year (not limited to Nature Index journals).

Technical terms

Active learning: A machine learning strategy where the model iteratively selects the most informative examples for human annotation to improve performance with fewer labelled data.

Natural language processing: Computational techniques for analysing and understanding human language, enabling automated parsing, classification and extraction of textual information.

Zero-shot classification: A method that assigns labels to data instances without direct training on those specific classes, relying instead on semantic representations or descriptions of tasks.

Systematic review: A structured and transparent approach to identifying, appraising and synthesising all relevant research on a specific question to minimise bias and improve reproducibility.

References

1. A living critical interpretive synthesis to yield a framework on the production and dissemination of living evidence syntheses for decision-making. Implementation Science (2024).
2. Automated literature research and review-generation method based on large language models. National Science Review (2025).
3. A novel application of machine learning and zero-shot classification methods for automated abstract screening in systematic reviews. Decision Analytics Journal (2023).
4. An open source machine learning framework for efficient and transparent systematic reviews. Nature Machine Intelligence (2021).
5. Software tools to support title and abstract screening for systematic reviews in healthcare: an evaluation. BMC Medical Research Methodology (2020).