scite: a smart citation index that displays the context of citations and classifies their intent using deep learning

Scientometrics, Bibliometrics, and Deep Learning

This preprint presents scite, a novel citation indexer that uses deep learning to classify citations based on their context within citing papers. This work falls under the normal science stage of Kuhn's paradigm cycle. The preprint operates within the existing paradigms of scientometrics and bibliometrics, seeking to refine and improve citation indexing rather than proposing a radical shift in how scientific impact is measured. While the use of deep learning is a relatively new technique in this field, it is applied here as a tool within the existing paradigm, not as a challenge to it. The authors acknowledge the limitations of existing citation indices and aim to address them by providing a more nuanced and informative approach to understanding citation context, thereby contributing to the ongoing refinement and puzzle-solving activity characteristic of normal science. They are working within the accepted framework of citation analysis, seeking to improve the accuracy and depth of information available, rather than overturning existing methodologies or theoretical foundations.