@inproceedings{juraska-walker-2021-attention,
title = "Attention Is Indeed All You Need: Semantically Attention-Guided Decoding for Data-to-Text {NLG}",
author = "Juraska, Juraj and
Walker, Marilyn",
editor = "Belz, Anya and
Fan, Angela and
Reiter, Ehud and
Sripada, Yaji",
booktitle = "Proceedings of the 14th International Conference on Natural Language Generation",
month = aug,
year = "2021",
address = "Aberdeen, Scotland, UK",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.inlg-1.45",
doi = "10.18653/v1/2021.inlg-1.45",
pages = "416--431",
abstract = "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.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="juraska-walker-2021-attention">
<titleInfo>
<title>Attention Is Indeed All You Need: Semantically Attention-Guided Decoding for Data-to-Text NLG</title>
</titleInfo>
<name type="personal">
<namePart type="given">Juraj</namePart>
<namePart type="family">Juraska</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marilyn</namePart>
<namePart type="family">Walker</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 14th International Conference on Natural Language Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anya</namePart>
<namePart type="family">Belz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Angela</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ehud</namePart>
<namePart type="family">Reiter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaji</namePart>
<namePart type="family">Sripada</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Aberdeen, Scotland, UK</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>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.</abstract>
<identifier type="citekey">juraska-walker-2021-attention</identifier>
<identifier type="doi">10.18653/v1/2021.inlg-1.45</identifier>
<location>
<url>https://aclanthology.org/2021.inlg-1.45</url>
</location>
<part>
<date>2021-08</date>
<extent unit="page">
<start>416</start>
<end>431</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Attention Is Indeed All You Need: Semantically Attention-Guided Decoding for Data-to-Text NLG
%A Juraska, Juraj
%A Walker, Marilyn
%Y Belz, Anya
%Y Fan, Angela
%Y Reiter, Ehud
%Y Sripada, Yaji
%S Proceedings of the 14th International Conference on Natural Language Generation
%D 2021
%8 August
%I Association for Computational Linguistics
%C Aberdeen, Scotland, UK
%F juraska-walker-2021-attention
%X 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.
%R 10.18653/v1/2021.inlg-1.45
%U https://aclanthology.org/2021.inlg-1.45
%U https://doi.org/10.18653/v1/2021.inlg-1.45
%P 416-431
Markdown (Informal)
[Attention Is Indeed All You Need: Semantically Attention-Guided Decoding for Data-to-Text NLG](https://aclanthology.org/2021.inlg-1.45) (Juraska & Walker, INLG 2021)
ACL