Generating News Headlines with Recurrent Neural Networks
This work addresses the task of automated headline generation for news articles, presenting an incremental improvement with a simplified attention mechanism.
The authors tackled the problem of generating news headlines from article text using an encoder-decoder RNN with LSTM and attention, finding it effective at concise paraphrasing. They also studied attention mechanisms, showing that a simplified version outperformed a more complex one on a held-out set.
We describe an application of an encoder-decoder recurrent neural network with LSTM units and attention to generating headlines from the text of news articles. We find that the model is quite effective at concisely paraphrasing news articles. Furthermore, we study how the neural network decides which input words to pay attention to, and specifically we identify the function of the different neurons in a simplified attention mechanism. Interestingly, our simplified attention mechanism performs better that the more complex attention mechanism on a held out set of articles.