Evaluate also the performances of the NER models used as base

[pafonta]

🚀 Feature

Evaluate also on our test sets the performances of the models used as base for training our NER models.

Motivation

We don't know if a change in the performances of our NER models is due to a change in the ones of the base models.

Besides, we could use these performances as baselines to measure the improvements we do.

Pitch

The base models from scipaCy used for training our NER models are evaluated as part of pipeline/ner/dvc.yaml.

One could compare the entity-level F1 scores between the base models and our models.

[pafonta]

@FrancescoCasalegno mentioned this is related to #276.