<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="matsubara-etal-2022-ensemble">
    <titleInfo>
        <title>Ensemble Transformer for Efficient and Accurate Ranking Tasks: an Application to Question Answering Systems</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yoshitomo</namePart>
        <namePart type="family">Matsubara</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Luca</namePart>
        <namePart type="family">Soldaini</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Eric</namePart>
        <namePart type="family">Lind</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Alessandro</namePart>
        <namePart type="family">Moschitti</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2022-12</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: EMNLP 2022</title>
        </titleInfo>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Abu Dhabi, United Arab Emirates</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Large transformer models can highly improve Answer Sentence Selection (AS2) tasks, but their high computational costs prevent their use in many real-world applications. In this paper, we explore the following research question: How can we make the AS2 models more accurate without significantly increasing their model complexity? To address the question, we propose a Multiple Heads Student architecture (named CERBERUS), an efficient neural network designed to distill an ensemble of large transformers into a single smaller model. CERBERUS consists of two components: a stack of transformer layers that is used to encode inputs, and a set of ranking heads; unlike traditional distillation technique, each of them is trained by distilling a different large transformer architecture in a way that preserves the diversity of the ensemble members. The resulting model captures the knowledge of heterogeneous transformer models by using just a few extra parameters. We show the effectiveness of CERBERUS on three English datasets for AS2; our proposed approach outperforms all single-model distillations we consider, rivaling the state-of-the-art large AS2 models that have 2.7\times more parameters and run 2.5\times slower. Code for our model is available at https://github.com/amazon-research/wqa-cerberus.</abstract>
    <identifier type="citekey">matsubara-etal-2022-ensemble</identifier>
    <location>
        <url>https://aclanthology.org/2022.findings-emnlp.537</url>
    </location>
    <part>
        <date>2022-12</date>
        <extent unit="page">
            <start>7259</start>
            <end>7272</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Ensemble Transformer for Efficient and Accurate Ranking Tasks: an Application to Question Answering Systems
%A Matsubara, Yoshitomo
%A Soldaini, Luca
%A Lind, Eric
%A Moschitti, Alessandro
%S Findings of the Association for Computational Linguistics: EMNLP 2022
%D 2022
%8 December
%I Association for Computational Linguistics
%C Abu Dhabi, United Arab Emirates
%F matsubara-etal-2022-ensemble
%X Large transformer models can highly improve Answer Sentence Selection (AS2) tasks, but their high computational costs prevent their use in many real-world applications. In this paper, we explore the following research question: How can we make the AS2 models more accurate without significantly increasing their model complexity? To address the question, we propose a Multiple Heads Student architecture (named CERBERUS), an efficient neural network designed to distill an ensemble of large transformers into a single smaller model. CERBERUS consists of two components: a stack of transformer layers that is used to encode inputs, and a set of ranking heads; unlike traditional distillation technique, each of them is trained by distilling a different large transformer architecture in a way that preserves the diversity of the ensemble members. The resulting model captures the knowledge of heterogeneous transformer models by using just a few extra parameters. We show the effectiveness of CERBERUS on three English datasets for AS2; our proposed approach outperforms all single-model distillations we consider, rivaling the state-of-the-art large AS2 models that have 2.7\times more parameters and run 2.5\times slower. Code for our model is available at https://github.com/amazon-research/wqa-cerberus.
%U https://aclanthology.org/2022.findings-emnlp.537
%P 7259-7272

Markdown (Informal)

[Ensemble Transformer for Efficient and Accurate Ranking Tasks: an Application to Question Answering Systems](https://aclanthology.org/2022.findings-emnlp.537) (Matsubara et al., Findings 2022)

• Ensemble Transformer for Efficient and Accurate Ranking Tasks: an Application to Question Answering Systems (Matsubara et al., Findings 2022)

ACL

• Yoshitomo Matsubara, Luca Soldaini, Eric Lind, and Alessandro Moschitti. 2022. Ensemble Transformer for Efficient and Accurate Ranking Tasks: an Application to Question Answering Systems. In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 7259–7272, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.