Global Encoding for Abstractive Summarization
This addresses issues in abstractive summarization for NLP applications, but it is incremental as it builds on existing seq2seq models.
The authors tackled the problem of repetition and semantic irrelevance in neural abstractive summarization by proposing a global encoding framework, which improved performance on LCSTS and English Gigaword datasets, outperforming baseline models and reducing repetition.
In neural abstractive summarization, the conventional sequence-to-sequence (seq2seq) model often suffers from repetition and semantic irrelevance. To tackle the problem, we propose a global encoding framework, which controls the information flow from the encoder to the decoder based on the global information of the source context. It consists of a convolutional gated unit to perform global encoding to improve the representations of the source-side information. Evaluations on the LCSTS and the English Gigaword both demonstrate that our model outperforms the baseline models, and the analysis shows that our model is capable of reducing repetition.