metadata

library_name: zeroshot_classifier
tags:
  - transformers
  - sentence-transformers
  - zeroshot_classifier
license: mit
datasets:
  - claritylab/UTCD
language:
  - en
pipeline_tag: zero-shot-classification
metrics:
  - accuracy

Zero-shot Implicit Binary BERT

This is a BERT model. It was introduced in the Findings of ACL'23 Paper Label Agnostic Pre-training for Zero-shot Text Classification by Christopher Clarke, Yuzhao Heng, Yiping Kang, Krisztian Flautner, Lingjia Tang and Jason Mars. The code for training and evaluating this model can be found here.

Model description

This model is intended for zero-shot text classification. It was trained under the binary classification framework via implicit training with the aspect-normalized UTCD dataset.

Finetuned from model: bert-base-uncased

Usage

Install our python package:

pip install zeroshot-classifier

Then, you can use the model like this:

>>> from zeroshot_classifier.models import BinaryBertCrossEncoder
>>> model = BinaryBertCrossEncoder(model_name='claritylab/zero-shot-implicit-binary-bert')

>>> text = "I'd like to have this track onto my Classical Relaxations playlist."
>>> labels = [
>>>     'Add To Playlist', 'Book Restaurant', 'Get Weather', 'Play Music', 'Rate Book', 'Search Creative Work',
>>>     'Search Screening Event'
>>> ]
>>> aspect = 'intent'
>>> aspect_sep_token = model.tokenizer.additional_special_tokens[0]
>>> text = f'{aspect} {aspect_sep_token} {text}'

>>> query = [[text, lb] for lb in labels]
>>> logits = model.predict(query, apply_softmax=True)
>>> print(logits)

[[7.3497969e-04 9.9926502e-01]
 [9.9988127e-01 1.1870124e-04]
 [9.9988961e-01 1.1033980e-04]
 [1.9227572e-03 9.9807727e-01]
 [9.9985313e-01 1.4685343e-04]
 [9.9938977e-01 6.1021477e-04]
 [9.9838030e-01 1.6197052e-03]]