LaBSE-Malach-Multilabel

A multilabel text classification model fine-tuned on the Visual History Archive in 6 languages. Training text segments consisted of ~350 words on average.

Given an input string, the model predicts probabilities for 2800 subject keyword IDs from the VHA thesaurus.

The VHA thesaurus, which maps IDs to textual labels and describes the hierarchical structure of the keywords, is not included in this repository.

Downloads last month
34
Safetensors
Model size
0.5B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for ufal/labse-malach-multilabel

Finetuned
(106)
this model