@inproceedings{huang-etal-2020-improve,
title = "Improve Transformer Models with Better Relative Position Embeddings",
author = "Huang, Zhiheng and
Liang, Davis and
Xu, Peng and
Xiang, Bing",
editor = "Cohn, Trevor and
He, Yulan and
Liu, Yang",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2020",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2020.findings-emnlp.298/",
doi = "10.18653/v1/2020.findings-emnlp.298",
pages = "3327--3335",
abstract = "The transformer model has demonstrated superior results on NLP tasks including machine translation and question answering. In this paper, we argue that the position information is not fully utilized in existing work. For example, the initial proposal of a sinusoid embedding is fixed and not learnable. In this paper, we first review the absolute position embeddings and existing relative position embedding methods. We then propose new methods to encourage increased interaction between query, key and relative position embeddings in the self-attention mechanism. Our most promising approach is a generalization of the absolute position embedding. Our method results in increased accuracy compared to previous approaches in absolute and relative position embeddings on the SQuAD1.1 dataset. In addition, we address the inductive property of whether a position embedding can be robust enough to handle long sequences. We demonstrate empirically that our relative embedding method can be reasonably generalized to and is robust in the inductive perspective. Finally, we show that our proposed method can be effectively and efficiently adopted as a near drop-in replacement for improving the accuracy of large models with little computational overhead."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2020-improve">
<titleInfo>
<title>Improve Transformer Models with Better Relative Position Embeddings</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhiheng</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Davis</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bing</namePart>
<namePart type="family">Xiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2020-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2020</title>
</titleInfo>
<name type="personal">
<namePart type="given">Trevor</namePart>
<namePart type="family">Cohn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yulan</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The transformer model has demonstrated superior results on NLP tasks including machine translation and question answering. In this paper, we argue that the position information is not fully utilized in existing work. For example, the initial proposal of a sinusoid embedding is fixed and not learnable. In this paper, we first review the absolute position embeddings and existing relative position embedding methods. We then propose new methods to encourage increased interaction between query, key and relative position embeddings in the self-attention mechanism. Our most promising approach is a generalization of the absolute position embedding. Our method results in increased accuracy compared to previous approaches in absolute and relative position embeddings on the SQuAD1.1 dataset. In addition, we address the inductive property of whether a position embedding can be robust enough to handle long sequences. We demonstrate empirically that our relative embedding method can be reasonably generalized to and is robust in the inductive perspective. Finally, we show that our proposed method can be effectively and efficiently adopted as a near drop-in replacement for improving the accuracy of large models with little computational overhead.</abstract>
<identifier type="citekey">huang-etal-2020-improve</identifier>
<identifier type="doi">10.18653/v1/2020.findings-emnlp.298</identifier>
<location>
<url>https://aclanthology.org/2020.findings-emnlp.298/</url>
</location>
<part>
<date>2020-11</date>
<extent unit="page">
<start>3327</start>
<end>3335</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Improve Transformer Models with Better Relative Position Embeddings
%A Huang, Zhiheng
%A Liang, Davis
%A Xu, Peng
%A Xiang, Bing
%Y Cohn, Trevor
%Y He, Yulan
%Y Liu, Yang
%S Findings of the Association for Computational Linguistics: EMNLP 2020
%D 2020
%8 November
%I Association for Computational Linguistics
%C Online
%F huang-etal-2020-improve
%X The transformer model has demonstrated superior results on NLP tasks including machine translation and question answering. In this paper, we argue that the position information is not fully utilized in existing work. For example, the initial proposal of a sinusoid embedding is fixed and not learnable. In this paper, we first review the absolute position embeddings and existing relative position embedding methods. We then propose new methods to encourage increased interaction between query, key and relative position embeddings in the self-attention mechanism. Our most promising approach is a generalization of the absolute position embedding. Our method results in increased accuracy compared to previous approaches in absolute and relative position embeddings on the SQuAD1.1 dataset. In addition, we address the inductive property of whether a position embedding can be robust enough to handle long sequences. We demonstrate empirically that our relative embedding method can be reasonably generalized to and is robust in the inductive perspective. Finally, we show that our proposed method can be effectively and efficiently adopted as a near drop-in replacement for improving the accuracy of large models with little computational overhead.
%R 10.18653/v1/2020.findings-emnlp.298
%U https://aclanthology.org/2020.findings-emnlp.298/
%U https://doi.org/10.18653/v1/2020.findings-emnlp.298
%P 3327-3335
Markdown (Informal)
[Improve Transformer Models with Better Relative Position Embeddings](https://aclanthology.org/2020.findings-emnlp.298/) (Huang et al., Findings 2020)
ACL