As the world is gearing towards model driven approach for most decision making, it is important to understand how a machine learning model is reaching to a decision. Most NLP models operate over high number of dimensions and so interpretability becomes even more important in such cases.
This repository is an effort to explain use of LIME in explaining a simple logistic alrogithm based text classification model. I will keep on adding more models in future to improve the scope of this analysis and to provide a good reference for any relevant work.
Data is taken from Jigsaw Toxic Comment Classification challenge of Kaggle and can be download from here. This problem was a multi-label problem but I changed it to binary classification problem for ease of explaination of generated results.
Explaination Chart
