SciPred: An end to end approach to classify predatory journals

In today's ruthless academic world, researchers are constantly under pressure to publish work. This has led to significant growth in the number of predatory publishers. With a questionable peer-review process, these publishing houses charge high processing fees in exchange for publishing articles. Given the heterogeneity in academia, identifying such articles is a herculean task. While conventional approaches use metadata of articles, recently an attempt has also been made to use the text itself to filter out such articles. We propose an end-to-end neural network-based pipeline to classify predatory journals. We use a higher dimensional SciBERT model and fine-tune it to the dataset. We also augment the dataset by data from other similar journals.