Attention Is Indeed All You Need: Semantically Attention-Guided Decoding for Data-to-Text NLG
This addresses the issue of unreliable information mention in generated text for data-to-text NLG, though it is incremental as it builds on existing encoder-decoder models like T5 and BART.
The paper tackles the problem of semantic inaccuracy in data-to-text generation by proposing a novel decoding method that uses cross-attention to guide beam search, resulting in dramatic reductions in semantic errors while maintaining state-of-the-art quality on three datasets.
Ever since neural models were adopted in data-to-text language generation, they have invariably been reliant on extrinsic components to improve their semantic accuracy, because the models normally do not exhibit the ability to generate text that reliably mentions all of the information provided in the input. In this paper, we propose a novel decoding method that extracts interpretable information from encoder-decoder models' cross-attention, and uses it to infer which attributes are mentioned in the generated text, which is subsequently used to rescore beam hypotheses. Using this decoding method with T5 and BART, we show on three datasets its ability to dramatically reduce semantic errors in the generated outputs, while maintaining their state-of-the-art quality.