CNN for sentences

This is my first time implementing a complete neural network myself and I am finding some difficulties.
I am implementing a paper’s neural network that predicts a score of an essay.
I finished the first part which is getting the essays embeddings.
Now I have an essays input of shape (batch_size, max_sent, max_word, embedding_dim)
max_sent: number of sentences in the longest essay
max_word: number of words in the longest sentence

Now I need to pass each sentence to a 1d CNN layer
then the output for each sentence is passed to its own attention pooling layer
I am unsure how to reshape my input in order to perform this operation.

If made a reply earlier today that might help.

Thanks but the architecture of the input I mentioned is different.