Closed BLKSerene closed 1 year ago
Actualy, sentiment
requires scikit-learn
as dependency.
joblib.load
loads a model, which is a sklearn.svm.LinearSVC
object
I love the idea underthesea
has least dependencies. Oneway to do that is we must implemnt SVM ourself, but a bigger issue come is to reinvent the wheel.
Similar to #505, I only use
sent_tokenize
,word_tokenize
,pos_tag
, andsentiment
for Vietnamese NLP tasks in my project. I have tested that these functinos do not depend onscikit-learn
, which is a large dependency.So I'm wondering that whether it is possible to also move
scikit-learn
toextras_require
as done in #506 so that lite users ofunderthesea
do not have to install such a large 3rd-party library?