: Vocabulaire du TAL: true negative rate

NLP resources and evaluation > measure > confusion matrix > true negative rate

true negative rate

Proportion of actual negatives that are correctly identified as such by a binary classification test. (Wikidata)

More specifically we can measure the True Positive Rate (TPR) and the True Negative Rate (TNR) on these subgroups which in turn inform equalized odds a standard metric used in the fairness literature. (Zhang, van de Meent & Wallace, 2021)
Sensitivity (recall) was 87.5% and specificity (true negative rates) was 94.1%. (Mendonca, Haas, Shagina, Larson & Friedman, 2003)
Their true positive rates and true negative rates are more balanced than human evaluators. (Garbacea, Carton, Yan & Mei, 2019)
We define specificity as the distance in levels between a given word and each concept found above it in the hierarchy. (Kennedy & Szpakowicz, 2008)
While the diagnostic probes drastically outperformed their controlled counterparts we do see a trend of selectivity improving with the number of layers. (Shapiro, Paullada & Steinert-Threlkeld, 2021)

http://data.loterre.fr/ark:/67375/8LP-R07M22T1-T

RDF/XML TURTLE JSON-LD Dernière modification le 23/10/2024

Loterre