This is the link.
In this paper, the authors used attention mechanizm to decide where to focus when decoder outputs.
This neural network model is a modification of the state-of-the-art machine translation model.
This method outperformed the current text abstraction methods.
These days, the improvements of RNN method are heavily affecting Natural Language Processing areas.
I think abstractive sentece summarization is one of the promising areas for advancement of NLP.