I have done text classification using LSTM architecture, the problem with this architecture is it overfits

Question

LSTM architecture to classify text within 30 categories, problem is it over fitting after 3 epochs,any suggestions to avoid over fitting ?

model.add(Embedding