The study reviews the State-of-the-Art datasets and solutions for automatic fact-checking and tested their applicability in production environments. Authors of the publication discovered overfitting issues in those models, and proposed a data filtering method that improves the model’s performance and generalization. Then, the scientists designed an unsupervised fine-tuning of the Masked Language models to improve its accuracy working with Wikipedia.
Authors also proposed a novel query enhancing method to improve evidence discovery using the Wikipedia Search API. Finally, the paper presents a new fact-checking system, the WikiCheck API that automatically performs a facts validation process based on the Wikipedia knowledge base.
Original link to publication: https://dl.acm.org/doi/abs/10.1145/3459637.3481961
Full version of publication: https://arxiv.org/pdf/2109.00835.pdf
Leave a Reply