Papers by Nadja Geisler

Copied to clipboard

Challenge:	Abstractive summarization methods require large corpora to train neural architectures.
Approach:	They propose a novel automatic corpus construction approach that automatically constructs large open-licensed summarization corpora from existing large text collections and an evaluation process with human annotators.
Outcome:	The proposed approach can be used to train abstractive summarization models on large corpora and through a manual evaluation with human annotators.