Abstract: Highlights•We propose a contrastive learning method for abstractive text summarization.•We construct contrastive samples in the source sentences.•We incorporate the salient distance-aware weightsinto the contrastive loss.•Experiments show that our model has obviously better performance.
Loading