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

“…(2021) found that just 0.31% of all citations in their dataset were instances of disagreement, a share that has remained relatively stable over time. This is consistent with figures from similar assessments of other scientific fields, which found 2.4% “negative” citations (Catalini et al., 2015) and 0.8% “disputing” citations (Nicholson et al., 2021) in their respective bibliographic datasets.…”

“…This is broadly in alignment with scite's overall assessments, as the creators indicate that "the average distribution of citation statements [is] 92.6% mentioning, 6.5% supporting, and 0.8% contrasting statements." 5 While this may be in alignment with scite's overall assessment of literature, it diverged significantly from our assessment of meaning, as we determined that, of the 96 citations scite classified as mentioning, 40 were more appropriately classified as supporting and 17 were more appropriately classified as contrasting. While scite determined that the majority of citations should be classified as mentioning, human assessors disagreed and had higher rates of supporting and contrasting citations.…”

“…The ‘Citations and impact’ section displays traditional citations and alternative metrics. The number of supporting or contrasting citations available via the ‘scite’ tool [ 55 ] is shown where possible. Users can also find data citations to the preprint, and the number of preprint recommendations by various groups in this section.…”