Highlight the words responsible for particular emotion predictions of Deep learning model

Question

I have created a Bidirectional LSTM model to predict sentiment of a sentence with one layer as Glove embedding. I want to use this model to highlight the words which are res