This is a project which contains different models that are built from stratch without using any inbuilt models and methods , that tries to do text classification . I am using amazon food reviews dataset here , and the labels I am considering for reviews are
1)positive
2)neutral
3)negative.
I have implemented several models like rule based classifier, naive bayes classifier and many models yet to come.
Link to dataset : https://www.kaggle.com/datasets/snap/amazon-fine-food-reviews
The test set accuracy are also mentioned at the the of corresponding notebooks. And note that these are just basic implementations and with many advanced practices , test accuracy can be improved further.
rule based classifier accuracy : around 44%
naive bayes classifier accuracy : around 78%