#bag-of-words

Question 1: Select all that apply

Which statements about encoding categorical and text features are true? Select all that apply. One of them refers to this code:

pipe = make_pipeline(CountVectorizer(), SVC())
cross_validate(pipe, X_train, y_train)

When a one-hot encoder represents unseen categories as all-zero rows (handle_unknown="ignore"), all unknown categories look the same to the model1
Raising max_features of a bag-of-words vectorizer (keeping more words) is likely to raise the training score2
A bag-of-words vectorizer fitted on the training split raises an error when a validation document contains a word it has never seen3
When this pipeline is cross-validated, different folds can end up with slightly different numbers of features4