Fig. 2
From: Estimating the prevalence of text overlap in biomedical conference abstracts
Pairs of similar abstracts are not trivially related to the number of abstracts within those pairs. Left, 12 abstracts can form 6 highly similar pairs. Right, 4 abstracts can form 6 highly similar pairs. In our work we used eTBLAST to make 1,235,085 pairwise comparisons. Approximately 2% (24,365 total) of those are highly similar pairs. Abstracts within highly similar pairs number 20,857. Apparently, there is non-trivial structure within the highly similar pairs of abstracts; however, assessing this structure is beyond the scope this work