Do Transformer Attention Heads Provide Transparency in Abstractive Summarization?

Open in new window