With growing global concern over environmental issues, the British Columbia Environmental Film Festival (BCEFF) seeks to present films with scientifically grounded stories that inspire action and change. However, its existing fact-checking process relied on a fully manual review. This project developed a RoBERTa-base claim detection model to reduce reviewer workload. The final model achieved an accuracy of 91%, with high recall and precision scores of 73. 3% and 75. 9%, demonstrating strong performance for first-pass screening. An environmental impact analysis estimated a low total training cost of 1. 53 Wh, reinforcing the project’s commitment to sustainability and justifying its implementation for environmental films. BCEFF’s reviewer workflow was updated alongside a user manual to integrate the model into the overall film screening process. Funding of CA9, 880 from the UBC Centre for Community Engaged Learning (Chapman and Innovation Grant) was obtained to support long-term operation and growth. Overall, this project demonstrates that a semi-automated fact-checking framework can improve review efficiency while preserving scientific standards, creating opportunities for adoption by film festivals beyond BCEFF.
Lee et al. (Thu,) studied this question.