<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chronis-erk-2020-bishop">
    <titleInfo>
        <title>When is a bishop not like a rook? When it’s like a rabbi! Multi-prototype BERT embeddings for estimating semantic relationships</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Gabriella</namePart>
        <namePart type="family">Chronis</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Katrin</namePart>
        <namePart type="family">Erk</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>Proceedings of the 24th Conference on Computational Natural Language Learning</title>
        </titleInfo>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Online</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>This paper investigates contextual language models, which produce token representations, as a resource for lexical semantics at the word or type level. We construct multi-prototype word embeddings from bert-base-uncased (Devlin et al., 2018). These embeddings retain contextual knowledge that is critical for some type-level tasks, while being less cumbersome and less subject to outlier effects than exemplar models. Similarity and relatedness estimation, both type-level tasks, benefit from this contextual knowledge, indicating the context-sensitivity of these processes. BERT’s token level knowledge also allows the testing of a type-level hypothesis about lexical abstractness, demonstrating the relationship between token-level phenomena and type-level concreteness ratings. Our findings provide important insight into the interpretability of BERT: layer 7 approximates semantic similarity, while the final layer (11) approximates relatedness.</abstract>
    <identifier type="citekey">chronis-erk-2020-bishop</identifier>
    <identifier type="doi">10.18653/v1/2020.conll-1.17</identifier>
    <location>
        <url>https://aclanthology.org/2020.conll-1.17</url>
    </location>
    <part>
        <date>2020-11</date>
        <extent unit="page">
            <start>227</start>
            <end>244</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T When is a bishop not like a rook? When it’s like a rabbi! Multi-prototype BERT embeddings for estimating semantic relationships
%A Chronis, Gabriella
%A Erk, Katrin
%S Proceedings of the 24th Conference on Computational Natural Language Learning
%D 2020
%8 November
%I Association for Computational Linguistics
%C Online
%F chronis-erk-2020-bishop
%X This paper investigates contextual language models, which produce token representations, as a resource for lexical semantics at the word or type level. We construct multi-prototype word embeddings from bert-base-uncased (Devlin et al., 2018). These embeddings retain contextual knowledge that is critical for some type-level tasks, while being less cumbersome and less subject to outlier effects than exemplar models. Similarity and relatedness estimation, both type-level tasks, benefit from this contextual knowledge, indicating the context-sensitivity of these processes. BERT’s token level knowledge also allows the testing of a type-level hypothesis about lexical abstractness, demonstrating the relationship between token-level phenomena and type-level concreteness ratings. Our findings provide important insight into the interpretability of BERT: layer 7 approximates semantic similarity, while the final layer (11) approximates relatedness.
%R 10.18653/v1/2020.conll-1.17
%U https://aclanthology.org/2020.conll-1.17
%U https://doi.org/10.18653/v1/2020.conll-1.17
%P 227-244

Markdown (Informal)

[When is a bishop not like a rook? When it’s like a rabbi! Multi-prototype BERT embeddings for estimating semantic relationships](https://aclanthology.org/2020.conll-1.17) (Chronis & Erk, CoNLL 2020)

• When is a bishop not like a rook? When it’s like a rabbi! Multi-prototype BERT embeddings for estimating semantic relationships (Chronis & Erk, CoNLL 2020)

ACL

• Gabriella Chronis and Katrin Erk. 2020. When is a bishop not like a rook? When it’s like a rabbi! Multi-prototype BERT embeddings for estimating semantic relationships. In Proceedings of the 24th Conference on Computational Natural Language Learning, pages 227–244, Online. Association for Computational Linguistics.