Extracting and quantifying eponyms in full-text articles

Eponyms are known to praise leading scientists for their contributions to science. Some are so widespread that they are even known by laypeople (e.g., Alzheimer's disease, Darwinism). However, there is no systematic way to discover the distributions of eponyms in scientific domains. Prior work has tackled this issue but has failed to address it completely. Early attempts involved the manual labelling of all eponyms found in a few textbooks of given domains, such as chemistry. Others relied on search engines to probe bibliographic records seeking a single eponym at a time, such as Nash Equilibrium. Nonetheless, we failed to find any attempt of eponym quantification in a large volume of full-text publications. This article introduces a semi-automatic text mining approach to extracting eponyms and quantifying their use in such datasets. Candidate eponyms are matched programmatically by regular expressions, and then validated manually. As a case study, the processing of 821 recent Scientometrics articles reveals a mixture of established and emerging eponyms. The results stress the value of text mining for the rapid extraction and quantification of eponyms that may have substantial implications for research evaluation