Firstly, you need to determine what the task you need (classification or clustering) and for what ( sentences, concepts, or relations).
Secondly, you need to select the text representation way to represent the text to Deep learning training and testing vectors. There’s many ways for text representation such as:
1- One Hot Encoding
2-Bag-of-Words(BOW)
3-Advanced BOW (TF-IDF-BOW)
4-Word2Vec
Then you need to select the satiable Deep learning Model (e.g. CNN, DBN, RNN…) , adjusting the parameters and do.
Deep Learning is based on using neural networks architectures. These models do not consume raw text but instead we need to convert the text into numerical vector values. The most widely used technique for this is Embeddings such as BERT, Word2Vect etc. Besides, you need to ensure that your text is cleaned and pre-processed before feeding the model. Here are some resources: