<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cao-etal-2020-deformer">
    <titleInfo>
        <title>DeFormer: Decomposing Pre-trained Transformers for Faster Question Answering</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Qingqing</namePart>
        <namePart type="family">Cao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Harsh</namePart>
        <namePart type="family">Trivedi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Aruna</namePart>
        <namePart type="family">Balasubramanian</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Niranjan</namePart>
        <namePart type="family">Balasubramanian</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2020-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Dan</namePart>
            <namePart type="family">Jurafsky</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Joyce</namePart>
            <namePart type="family">Chai</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Natalie</namePart>
            <namePart type="family">Schluter</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Joel</namePart>
            <namePart type="family">Tetreault</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>Transformer-based QA models use input-wide self-attention – i.e. across both the question and the input passage – at all layers, causing them to be slow and memory-intensive. It turns out that we can get by without input-wide self-attention at all layers, especially in the lower layers. We introduce DeFormer, a decomposed transformer, which substitutes the full self-attention with question-wide and passage-wide self-attentions in the lower layers. This allows for question-independent processing of the input text representations, which in turn enables pre-computing passage representations reducing runtime compute drastically. Furthermore, because DeFormer is largely similar to the original model, we can initialize DeFormer with the pre-training weights of a standard transformer, and directly fine-tune on the target QA dataset. We show DeFormer versions of BERT and XLNet can be used to speed up QA by over 4.3x and with simple distillation-based losses they incur only a 1% drop in accuracy. We open source the code at https://github.com/StonyBrookNLP/deformer.</abstract>
    <identifier type="citekey">cao-etal-2020-deformer</identifier>
    <identifier type="doi">10.18653/v1/2020.acl-main.411</identifier>
    <location>
        <url>https://aclanthology.org/2020.acl-main.411</url>
    </location>
    <part>
        <date>2020-07</date>
        <extent unit="page">
            <start>4487</start>
            <end>4497</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T DeFormer: Decomposing Pre-trained Transformers for Faster Question Answering
%A Cao, Qingqing
%A Trivedi, Harsh
%A Balasubramanian, Aruna
%A Balasubramanian, Niranjan
%Y Jurafsky, Dan
%Y Chai, Joyce
%Y Schluter, Natalie
%Y Tetreault, Joel
%S Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
%D 2020
%8 July
%I Association for Computational Linguistics
%C Online
%F cao-etal-2020-deformer
%X Transformer-based QA models use input-wide self-attention – i.e. across both the question and the input passage – at all layers, causing them to be slow and memory-intensive. It turns out that we can get by without input-wide self-attention at all layers, especially in the lower layers. We introduce DeFormer, a decomposed transformer, which substitutes the full self-attention with question-wide and passage-wide self-attentions in the lower layers. This allows for question-independent processing of the input text representations, which in turn enables pre-computing passage representations reducing runtime compute drastically. Furthermore, because DeFormer is largely similar to the original model, we can initialize DeFormer with the pre-training weights of a standard transformer, and directly fine-tune on the target QA dataset. We show DeFormer versions of BERT and XLNet can be used to speed up QA by over 4.3x and with simple distillation-based losses they incur only a 1% drop in accuracy. We open source the code at https://github.com/StonyBrookNLP/deformer.
%R 10.18653/v1/2020.acl-main.411
%U https://aclanthology.org/2020.acl-main.411
%U https://doi.org/10.18653/v1/2020.acl-main.411
%P 4487-4497

Markdown (Informal)

[DeFormer: Decomposing Pre-trained Transformers for Faster Question Answering](https://aclanthology.org/2020.acl-main.411) (Cao et al., ACL 2020)

• DeFormer: Decomposing Pre-trained Transformers for Faster Question Answering (Cao et al., ACL 2020)

ACL

• Qingqing Cao, Harsh Trivedi, Aruna Balasubramanian, and Niranjan Balasubramanian. 2020. DeFormer: Decomposing Pre-trained Transformers for Faster Question Answering. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4487–4497, Online. Association for Computational Linguistics.