@inproceedings{ferrando-etal-2023-explaining,
title = "Explaining How Transformers Use Context to Build Predictions",
author = "Ferrando, Javier and
G{\'a}llego, Gerard I. and
Tsiamas, Ioannis and
Costa-juss{\`a}, Marta R.",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.301",
doi = "10.18653/v1/2023.acl-long.301",
pages = "5486--5513",
abstract = "Language Generation Models produce words based on the previous context. Although existing methods offer input attributions as explanations for a model{'}s prediction, it is still unclear how prior words affect the model{'}s decision throughout the layers. In this work, we leverage recent advances in explainability of the Transformer and present a procedure to analyze models for language generation. Using contrastive examples, we compare the alignment of our explanations with evidence of the linguistic phenomena, and show that our method consistently aligns better than gradient-based and perturbation-based baselines. Then, we investigate the role of MLPs inside the Transformer and show that they learn features that help the model predict words that are grammatically acceptable. Lastly, we apply our method to Neural Machine Translation models, and demonstrate that they generate human-like source-target alignments for building predictions.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ferrando-etal-2023-explaining">
<titleInfo>
<title>Explaining How Transformers Use Context to Build Predictions</title>
</titleInfo>
<name type="personal">
<namePart type="given">Javier</namePart>
<namePart type="family">Ferrando</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gerard</namePart>
<namePart type="given">I</namePart>
<namePart type="family">Gállego</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ioannis</namePart>
<namePart type="family">Tsiamas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marta</namePart>
<namePart type="given">R</namePart>
<namePart type="family">Costa-jussà</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Language Generation Models produce words based on the previous context. Although existing methods offer input attributions as explanations for a model’s prediction, it is still unclear how prior words affect the model’s decision throughout the layers. In this work, we leverage recent advances in explainability of the Transformer and present a procedure to analyze models for language generation. Using contrastive examples, we compare the alignment of our explanations with evidence of the linguistic phenomena, and show that our method consistently aligns better than gradient-based and perturbation-based baselines. Then, we investigate the role of MLPs inside the Transformer and show that they learn features that help the model predict words that are grammatically acceptable. Lastly, we apply our method to Neural Machine Translation models, and demonstrate that they generate human-like source-target alignments for building predictions.</abstract>
<identifier type="citekey">ferrando-etal-2023-explaining</identifier>
<identifier type="doi">10.18653/v1/2023.acl-long.301</identifier>
<location>
<url>https://aclanthology.org/2023.acl-long.301</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>5486</start>
<end>5513</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Explaining How Transformers Use Context to Build Predictions
%A Ferrando, Javier
%A Gállego, Gerard I.
%A Tsiamas, Ioannis
%A Costa-jussà, Marta R.
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F ferrando-etal-2023-explaining
%X Language Generation Models produce words based on the previous context. Although existing methods offer input attributions as explanations for a model’s prediction, it is still unclear how prior words affect the model’s decision throughout the layers. In this work, we leverage recent advances in explainability of the Transformer and present a procedure to analyze models for language generation. Using contrastive examples, we compare the alignment of our explanations with evidence of the linguistic phenomena, and show that our method consistently aligns better than gradient-based and perturbation-based baselines. Then, we investigate the role of MLPs inside the Transformer and show that they learn features that help the model predict words that are grammatically acceptable. Lastly, we apply our method to Neural Machine Translation models, and demonstrate that they generate human-like source-target alignments for building predictions.
%R 10.18653/v1/2023.acl-long.301
%U https://aclanthology.org/2023.acl-long.301
%U https://doi.org/10.18653/v1/2023.acl-long.301
%P 5486-5513
Markdown (Informal)
[Explaining How Transformers Use Context to Build Predictions](https://aclanthology.org/2023.acl-long.301) (Ferrando et al., ACL 2023)
ACL
- Javier Ferrando, Gerard I. Gállego, Ioannis Tsiamas, and Marta R. Costa-jussà. 2023. Explaining How Transformers Use Context to Build Predictions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5486–5513, Toronto, Canada. Association for Computational Linguistics.