ASReview is an open-source machine learning framework designed to accelerate the process of screening large volumes of textual data for systematic reviews. It targets researchers and professionals across disciplines who need to efficiently identify relevant literature, aiming to reduce the number of records a human needs to read while minimizing false negatives.

How It Works

ASReview employs active learning to interactively query researchers, who act as "oracles" to label data. The system prioritizes records for review based on the model's uncertainty, aiming for maximum efficiency. It supports three modes: Oracle Screen for interactive labeling, Exploration for demonstrating with pre-labeled data, and Simulation for evaluating model performance on fully labeled datasets.

Quick Start & Requirements

• Install via pip: pip install asreview
• Requires Python 3.8 or later.
• Docker installation is also available.
• Detailed installation guides for Windows and macOS are provided.
• Official documentation: asreview.readthedocs.io
• Quick tour: ASReview LAB quick tour

Highlighted Details

• Implements active learning for efficient systematic reviews.
• Published in Nature Machine Intelligence.
• Offers three distinct modes: Oracle Screen, Exploration, and Simulation.
• Provides citation for both methodology and software releases.

Maintenance & Community

ASReview LAB is maintained by Jonathan de Bruin and Yongchao Terry Ma. Further contact and resources include:

• Documentation: asreview.readthedocs.io
• Newsletter: asreview.ai/newsletter/subscribe
• Issue tracker: ASReview issue tracker
• Discussions: ASReview Discussions

Licensing & Compatibility

The ASReview software is licensed under the Apache 2.0 License. This license permits commercial use and linking with closed-source projects.

Limitations & Caveats

The project disclaimer states that the ASReview team accepts no responsibility or liability for the use of the tool or any damages arising from its application.