<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="clark-etal-2022-canine">
    <titleInfo>
        <title>Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jonathan</namePart>
        <namePart type="given">H</namePart>
        <namePart type="family">Clark</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Dan</namePart>
        <namePart type="family">Garrette</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Iulia</namePart>
        <namePart type="family">Turc</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">John</namePart>
        <namePart type="family">Wieting</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2022</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <genre authority="bibutilsgt">journal article</genre>
    <relatedItem type="host">
        <titleInfo>
            <title>Transactions of the Association for Computational Linguistics</title>
        </titleInfo>
        <originInfo>
            <issuance>continuing</issuance>
            <publisher>MIT Press</publisher>
            <place>
                <placeTerm type="text">Cambridge, MA</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">periodical</genre>
        <genre authority="bibutilsgt">academic journal</genre>
    </relatedItem>
    <abstract>Pipelined NLP systems have largely been superseded by end-to-end neural modeling, yet nearly all commonly used models still require an explicit tokenization step. While recent tokenization approaches based on data-derived subword lexicons are less brittle than manually engineered tokenizers, these techniques are not equally suited to all languages, and the use of any fixed vocabulary may limit a model’s ability to adapt. In this paper, we present Canine, a neural encoder that operates directly on character sequences—without explicit tokenization or vocabulary—and a pre-training strategy that operates either directly on characters or optionally uses subwords as a soft inductive bias. To use its finer-grained input effectively and efficiently, Canine combines downsampling, which reduces the input sequence length, with a deep transformer stack, which encodes context. Canine outperforms a comparable mBert model by 5.7 F1 on TyDi QA, a challenging multilingual benchmark, despite having fewer model parameters.</abstract>
    <identifier type="citekey">clark-etal-2022-canine</identifier>
    <identifier type="doi">10.1162/tacl_a_00448</identifier>
    <location>
        <url>https://aclanthology.org/2022.tacl-1.5</url>
    </location>
    <part>
        <date>2022</date>
        <detail type="volume"><number>10</number></detail>
        <extent unit="page">
            <start>73</start>
            <end>91</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Journal Article
%T Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation
%A Clark, Jonathan H.
%A Garrette, Dan
%A Turc, Iulia
%A Wieting, John
%J Transactions of the Association for Computational Linguistics
%D 2022
%V 10
%I MIT Press
%C Cambridge, MA
%F clark-etal-2022-canine
%X Pipelined NLP systems have largely been superseded by end-to-end neural modeling, yet nearly all commonly used models still require an explicit tokenization step. While recent tokenization approaches based on data-derived subword lexicons are less brittle than manually engineered tokenizers, these techniques are not equally suited to all languages, and the use of any fixed vocabulary may limit a model’s ability to adapt. In this paper, we present Canine, a neural encoder that operates directly on character sequences—without explicit tokenization or vocabulary—and a pre-training strategy that operates either directly on characters or optionally uses subwords as a soft inductive bias. To use its finer-grained input effectively and efficiently, Canine combines downsampling, which reduces the input sequence length, with a deep transformer stack, which encodes context. Canine outperforms a comparable mBert model by 5.7 F1 on TyDi QA, a challenging multilingual benchmark, despite having fewer model parameters.
%R 10.1162/tacl_a_00448
%U https://aclanthology.org/2022.tacl-1.5
%U https://doi.org/10.1162/tacl_a_00448
%P 73-91

Markdown (Informal)

[Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://aclanthology.org/2022.tacl-1.5) (Clark et al., TACL 2022)

• Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation (Clark et al., TACL 2022)

ACL

• Jonathan H. Clark, Dan Garrette, Iulia Turc, and John Wieting. 2022. Canine: Pre-training an Efficient Tokenization-Free Encoder for Language Representation. Transactions of the Association for Computational Linguistics, 10:73–91.