A multilabel text classification model fine-tuned on a the Visual History Archive in 6 languages. Input text segments consisted of ~350 words on average.
Given an input string, the model predicts probablites for 2800 subject keyword IDs from the VHA ontology.
