Joined Special Interest Group Active Learning
- Email: firstname.lastname@example.orgPhone: 030 253 7982Professor
- Phone: 030 253 8101Professor
- Email: email@example.comPhone: 030 253 4183Associate Professor
- Email: firstname.lastname@example.orgAssistant Professor
- Jonathan de Bruin (ITS)
- Jan de Boer (UBU)
- Felix Weijdema (UBU)
- Bianca Kramer (UBU)
- Martijn Huijts (ITS)
With the emergence of online publishing, the number of scientific papers on any topic is skyrocketing. Simultaneously, the public press and social media also produce data by the second. In this tsunami of new knowledge, imagine updating a medical guideline, making evidence-based policy, or scouting for new technologies: there’s not enough time to read everything. More and more researchers and organizations rely upon Systematic Reviews: attempts to synthesize state-of-the-art in a particular field. But this process is under severe strain: the exponential growth in textual data means they too have to screen an ever-larger body of work – resulting in costly, abandoned, or error-prone work. The rapidly evolving field of Artificial Intelligence (AI) has allowed the development of AI-aided pipelines that assist in finding relevant texts for search tasks. A well-established approach to increase efficiency is screening prioritization via active learning: a constant interaction between a human and a machine.
The Joined Special Interest Group Active Learning (SIG-AL) covers all aspects of using Active Learning for screening prioritization. To this end, the four specific objectives of this SIG are to advance state of the art by:
- Developing and evaluating even better methods to improve efficiency: Even the most negligible improvements might result in hundreds of hours of screening time not being needed.
- Developing Explainable AI (XAI) to understand the output produced by the machine and, as such, gain trust in the use of AL.
- Designing methods to ‘standardize’ and FAIR-ify the input data and all data produced by the active learning models for complete reproducibility and transparency.
- Applying active learning to use-cases beneficial to the general public.
The SIG is a collaboration between the focus area Applied Data Science and Human-centered Artificial Intelligence.
The organization ASReview in Github contains many repositories useful for applying, testing, and evaluating Active Learning. Some examples are:
- The user-friendly front-end of ASReview for anyone who would like to use the Active Learning models to speed up the screening phase of systematic reviews.
- The simulation mode for comparing and validating active learning models.
- A data platform with many labeled datasets which can be used for validation studies.
- Use-cases like the CORD19 database, screening medical guidelines for the Royal Dutch Pharmacists Association.
The SIG-AL is an open community, and everyone can contribute, mainly via Github. We also keep each other up to date via our Slack channel (send us an email if you want to join the open channel), which is our primary way of communication. We also have meetings: weekly stand-up, presentations from team members and researchers from outside, and many smaller sub-meetings to discuss technical topics, user experience, GitHub, etc. One day a week, the University provides us with a room to work together at the University.
The ASReview team also facilitates bachelor and master thesis projects.
Contact us via our discussion board, or via Slack [receive an invitation via email@example.com].