ATIS is Intent classification dataset in the well known in the field of NLU. This repository contains the implementation of Explainable AI techniques on models predicting intent using ATIS dataset. Explainable AI techniques explored here are LIME , Anchors and Integraded gradiants for classifiers RandomForest and simple 2 layer LSTM neural network.
Some examples for model explainations are
LSTM LIME
LSTM Integraded Gradient
Link for the dataset is https://www.kaggle.com/datasets/hassanamin/atis-airlinetravelinformationsystem
For Explainable AI techniques refer to LIME :- https://arxiv.org/abs/1602.04938
Integraded gradiants :- https://arxiv.org/abs/1703.01365
Anchors :- https://homes.cs.washington.edu/~marcotcr/aaai18.pdf