My question relates directly to the text summarization task. I know that there are a bunch of implementations with RNN networks (especially LSTMs) that use sentence-level at
