AI for AI: Using AI methods for classifying AI science documents

Abstract Subject area classification is an important first phase in the entire process involved bibliometrics. In this paper, we explore possibility of using automated algorithms for classifying scientific papers related to Artificial Intelligence at document level. The current semimanual and journal based, a realization that, argue, opens up potential inaccuracies. To counter this, our proposed approach makes use neural networks, specifically BERT. accuracy model reaches 96.5%. addition, was used further documents from 26 different subject areas Scopus database. Our findings indicate that significant subset existing Computer Science, Decision Mathematics publications could potentially be classified as AI-related. same holds particular cases other science fields such Medicine Psychology or Arts Humanities. above processes, there room automatic approaches utilized complementary manner with traditional manual procedures.