MoritzLaurer / deberta-v3-base-zeroshot-v1

deberta-v3-base-zeroshot-v1

Model description

The model is designed for zero-shot classification with the Hugging Face pipeline. The model should be substantially better at zero-shot classification than my other zero-shot models on the Hugging Face hub: https://huggingface.co/MoritzLaurer .

The model can do one universal task: determine whether a hypothesis is true or not_true given a text (also called entailment vs. not_entailment ).
This task format is based on the Natural Language Inference task (NLI). The task is so universal that any classification task can be reformulated into the task.

Training data

The model was trained on a mixture of 27 tasks and 310 classes that have been reformatted into this universal format.

1. 26 classification tasks with ~400k texts: 'amazonpolarity', 'imdb', 'appreviews', 'yelpreviews', 'rottentomatoes', 'emotiondair', 'emocontext', 'empathetic', 'financialphrasebank', 'banking77', 'massive', 'wikitoxic_toxicaggregated', 'wikitoxic_obscene', 'wikitoxic_threat', 'wikitoxic_insult', 'wikitoxic_identityhate', 'hateoffensive', 'hatexplain', 'biasframes_offensive', 'biasframes_sex', 'biasframes_intent', 'agnews', 'yahootopics', 'trueteacher', 'spam', 'wellformedquery'. See details on each dataset here: https://docs.google.com/spreadsheets/d/1Z18tMh02IiWgh6o8pfoMiI_LH4IXpr78wd_nmNd5FaE/edit?usp=sharing
2. Five NLI datasets with ~885k texts: "mnli", "anli", "fever", "wanli", "ling"

Note that compared to other NLI models, this model predicts two classes ( entailment vs. not_entailment ) as opposed to three classes (entailment/neutral/contradiction)

How to use the model

Simple zero-shot classification pipeline

from transformers import pipeline
classifier = pipeline("zero-shot-classification", model="MoritzLaurer/deberta-v3-base-zeroshot-v1")
sequence_to_classify = "Angela Merkel is a politician in Germany and leader of the CDU"
candidate_labels = ["politics", "economy", "entertainment", "environment"]
output = classifier(sequence_to_classify, candidate_labels, multi_label=False)
print(output)

Details on data and training

The code for preparing the data and training & evaluating the model is fully open-source here: https://github.com/MoritzLaurer/zeroshot-classifier/tree/main

Limitations and bias

The model can only do text classification tasks.

Please consult the original DeBERTa paper and the papers for the different datasets for potential biases.

License

The base model (DeBERTa-v3) is published under the MIT license. The datasets the model was fine-tuned on are published under a diverse set of licenses. The following spreadsheet provides an overview of the non-NLI datasets used for fine-tuning. The spreadsheets contains information on licenses, the underlying papers etc.: https://docs.google.com/spreadsheets/d/1Z18tMh02IiWgh6o8pfoMiI_LH4IXpr78wd_nmNd5FaE/edit?usp=sharing

In addition, the model was also trained on the following NLI datasets: MNLI, ANLI, WANLI, LING-NLI, FEVER-NLI.

Citation

If you use this model, please cite:

@article{laurer_less_2023,
    title = {Less {Annotating}, {More} {Classifying}: {Addressing} the {Data} {Scarcity} {Issue} of {Supervised} {Machine} {Learning} with {Deep} {Transfer} {Learning} and {BERT}-{NLI}},
    issn = {1047-1987, 1476-4989},
    shorttitle = {Less {Annotating}, {More} {Classifying}},
    url = {https://www.cambridge.org/core/product/identifier/S1047198723000207/type/journal_article},
    doi = {10.1017/pan.2023.20},
    language = {en},
    urldate = {2023-06-20},
    journal = {Political Analysis},
    author = {Laurer, Moritz and Van Atteveldt, Wouter and Casas, Andreu and Welbers, Kasper},
    month = jun,
    year = {2023},
    pages = {1--33},
}

Ideas for cooperation or questions?

If you have questions or ideas for cooperation, contact me at m{dot}laurer{at}vu{dot}nl or LinkedIn

Debugging and issues

Note that DeBERTa-v3 was released on 06.12.21 and older versions of HF Transformers seem to have issues running the model (e.g. resulting in an issue with the tokenizer). Using Transformers>=4.13 might solve some issues.