Which Spacy components are required to for optimal Lemmatizer quality?

[joshdavham]

I'm a frequent user of the Spacy library, and in particular, I use it to create frequency lists of lemmas on large data sets. Effectively, all I'm doing is taking text and getting their lemmas.

However, when using the common

nlp = spacy.load('ja_core_news_sm')

the processing is too slow.

After reading the docs, I then tried

nlp = spacy.load('ja_core_news_sm', enable=['lemmatizer'])

which is certainly faster, but the quality of lemmatization dropped sharply.

Clearly the answer is to add enable more components to get better, lemmatization. Perhaps something like this?

nlp = spacy.load('ja_core_news_sm', enable=['tok2vec', 'morphologizer', 'tagger', 'attribute_ruler', 'lemmatizer'])

Anyways, my question is: Which Spacy compontents are required for optimal Lemmatizer quality? I'd like to disable every other component that doesn't directly contribute to better lemmatization. Unfortunately I haven't found anything in the docs :(

Thanks!