Lemmatization improvements

[lefnire]

I have CleanText.keywords_fast() and .keywords_accurate(), using lemminflect & Stanford NLP (spacy-stanza) respectively. I'm not confident on the setups, I feel I could be using both tools more effectively. Especially, I have a crap-ton of custom regex in the methods above, which I assume could be handled via nlp.pipe() more elegantly / robustly.

I'll start this thread for things as I think up.

• Do I need pip install -U spacy-lookups-data for lemmatization? (screenshot)
• Does current use of Lemminflect bypass ^, and not need that installation? Which produces better lemmas? (Do some testing b/w spacy-lookups, lemminflect, stanza)