SNSF Data Team

Based on a government mandate, the Swiss National Science Foundation (SNSF) supports scientific research in all academic disciplines. It is the leading organisation for the promotion of scientific research in Switzerland. On the Data Portal, the SNSF publishes data on the evaluated projects and the persons involved in order to provide transparency and facilitate the analysis of funding activities.

In this space, the SNSF Data Team provides fine-tuned models for classifying grant peer review texts along 12 categories relevant to the evaluation criteria specified by the SNSF. In particular, the models are based on the allenai/specter2_base model and fine-tuned for a binary classification task on a sentence level. For more details on the methods, see the the following paper published in Quantitative Science Studies:

A Supervised Machine Learning Approach for Assessing Grant Peer Review Reports

by Gabriel Okasa, Alberto de León, Michaela Strinzel, Anne Jorstad, Katrin Milzow, Matthias Egger, and Stefan Müller, available open-access: https://doi.org/10.1162/QSS.a.23 . The fine-tuning codes are open-sourced on GitHub: https://github.com/snsf-data/ml-peer-review-analysis .

The model cards provide further details on the models, the fine-tuning procedure and evaluation metrics as well as minimal examples for usage of the models.

A follow-up paper published in Plos One:

Gender and disciplinary differences in grant proposal peer review: Content and sentiment in 39,280 reports

by Stefan Müller Gabriel Okasa, Michaela Strinzel, Anne Jorstad, Katrin Milzow, and Matthias Egger, available open-access: https://doi.org/10.1371/journal.pone.0352900 applies the fine-tuned models on the grant peer review reports submitted to the SNSF between 2016 and 2023. The results show that the tone, length, and focus of peer review reports varied systematically by reviewer gender and disciplinary context.

If you are using the models, please cite:

• Okasa G, de León A, Strinzel M, Jorstad A, Milzow K, Egger M, & Müller S (2025) A supervised machine learning approach for assessing grant peer review reports. Quantitative Science Studies 6 1189–1214. https://doi.org/10.1162/QSS.a.23

If you are referring to the analysis of grant peer review reports, please cite:

• Müller S, Okasa G, Strinzel M, Jorstad A, Milzow K, Egger M (2026) Gender and disciplinary differences in grant proposal peer review: Content and sentiment in 39,280 reports. PLoS One 21(9): e0352900. https://doi.org/10.1371/journal.pone.0352900