Counting the Gap: corpus and gender assignments for Nigerian computing research authorship, 2015-2025

Abstract

Corpus, author-level gender assignments and analysis code accompanying the paper"Counting the Gap: Eleven Years of Stagnant Female Authorship in Nigerian ComputingResearch (2015-2025)", presented at CWIT 2026, Federal University of Agriculture,Abeokuta.The corpus is 4,101 journal articles dated 2015 to 2025, retrieved from the Crossref RESTAPI where an author affiliation string contains "Nigeria", crossed with ten computing termqueries and deduplicated by DOI. It carries 20,141 author mentions, of which 12,922 areNigerian-affiliated.Gender was assigned in two published steps and no hand-built list was used. Forenameswere looked up in the names-dataset reference, keeping assignments above 0.60probability; forenames the reference missed were classified by a character n-gram logisticregression trained on that same reference, five-fold cross-validated accuracy 0.808,keeping predictions above 0.80 probability. Combined coverage is 77.6% of Nigerianauthor mentions.IMPORTANT: all gender labels are inferred from forenames. No author was asked. Thelabels are binary because the method cannot be otherwise, which is a limitation of themethod and not a claim about gender, and any individual label may be wrong. These filessupport proportions over a corpus and should not be used to state anything about a namedperson. Unresolved names are excluded from every proportion and are never counted asmale.