Leveraging Hugging Face for complex text classification use cases

[YeonwooSung]

Leveraging Hugging Face for complex text classification use cases

Summary:

1. Use Sentence Transformers to calculate the sentence vector of the input sentence, rather than calculate vectors for each word.

2. HuggingFace suggested to use SetFit, which is a library that helps the custom finetuning process of the sentence-transformers

3. Both HuggingFace and Witty Works tested various sentence-transformers models with diverse classifiers, and they settled with mpnet-base-v2 with logistic regression and KNN classifier.