We propose a selective encoding model to extend the sequence-to-sequence framework for abstractive sentence summarization. It consists of a sentence encoder, a selective gate network, and an attention equipped decoder. The sentence encoder and decoder are built with recurrent neural networks. The selective gate network constructs a second level sentence representation by controlling the informa...

Previous work has demonstrated that end-to-end neural sequence models well for document-level event role filler extraction. However, the network model suffers from problem of not being able to utilize global information, resulting in incomplete extraction arguments. This is because inputs BiLSTM are all single-word vectors with no input contextual information. phenomenon particularly pronounced...

Fluent readers process text at a number of levels; at the word, the sentence, the conceptual and the topic levels (Munro & Munro, 1991, 1994). The outputs from each process are retained for varying durations in short-term working memory. Depending in part on their purposes for reading at any time, readers perceive the outputs of some levels to be more useful than those at other levels. As they ...