How does Naïve Bayes integrate with modern embedding-based NLP pipelines?

Naïve Bayes can be applied on top of embeddings or used as a baseline alongside deep NLP models.

While NB traditionally uses bag-of-words, it can be extended using discretized embeddings or feature transformations. However, embeddings reduce sparsity and violate independence assumptions more strongly. In practice, NB is often used as a fast baseline compared to transformer-based models.