This is an example of using tf.keras to build an encoder-decoder model with attention mechanism to generate the title from the abstract. If you want to get better performance, I suggest to augment the dataset, even build the pre-trained word embeddings from the specific domain.
title_parameter.py reserves the hyper-parameters
title_dataset.py builds data set from title.json
title_train.py trains a model and save it to a file
title-predict.py is an example for using the model to generate title from abstract