This is a SetFit model that can be used for text classification.
It was created to predict the gender of runaways advertised in Danish Newspapers from 1750–1850 as part of the project Run Away at the Department of Politics and Society, Aalborg University.
The model is designed to explore how gender was encoded in a specific historical text genre. The model was trained on a sample of 1490 advertisements tagged for gender by Sofus Landor Dam and Anders Dyrborg Birkemose.
The model has an accuracy of 0.9888143176733781. The base model is CALDISS-AAU/DA-BERT_Old_News_V1. While this is not a sentence transformer, but a fill-mask model, the model performs well enough to be useful. The data was split 0.3 for testing.
Getting started
from setfit import SetFitModel
model = SetFitModel.from_pretrained("JohanHeinsen/Runaway_Advertisements_Gender")
preds = model(["Min tjenestepige løb væk fra mig i nat", "Soldaten Jonas er forsvundet fra mit hus."])
label_map = {0: "mand", 1: "kvinde"}
predicted_labels = [label_map[int(preds[0])], label_map[int(preds[1])]]
predicted_labels
