Content-Categerization-Library

The Pipeline uses different libraries and techniques manually implemented in python to handle multi-class classification problems which promises higher accuracy than the native TF-IDF feature extraction pipeline. The pipeline can be used to get baseline scores for text categorisation.

Algorithms used for Pipeline:

Filtering --

Zipf Law and Chi-square filters

Feature Extraction --

TF-IDF, Bag Of Words, GLOVE vectors , Novel document vector aggregation using GloVe vectors

Modelling --

Tree based algorithms -->

Light GBM, XGboost, Random Forest

Linear Algorithms -->

Logistic, SGD classifier

The pipeline Outputs a classification score report of all individual classes with corresponding weighted Precision and Recall.