<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rastegar-ramezani-2024-greek">
    <titleInfo>
        <title>From ‘It‘s All Greek to Me’ to ‘Nur Bahnhof Verstehen’: An Investigation of mBERT‘s Cross-Linguistic Capabilities</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Aria</namePart>
        <namePart type="family">Rastegar</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Pegah</namePart>
        <namePart type="family">Ramezani</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2024-12</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Felice</namePart>
            <namePart type="family">Dell’Orletta</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Alessandro</namePart>
            <namePart type="family">Lenci</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Simonetta</namePart>
            <namePart type="family">Montemagni</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Rachele</namePart>
            <namePart type="family">Sprugnoli</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>CEUR Workshop Proceedings</publisher>
            <place>
                <placeTerm type="text">Pisa, Italy</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-12-210-7060-6</identifier>
    </relatedItem>
    <abstract>This study investigates the impact of cross-linguistic similarities on idiom representation in mBERT, focusing on English and German idioms categorized by different degrees of similarity. We aim to determine whether different degrees of cross-linguistic similarities significantly affect mBERT‘s representations and to observe how these representations change across its 12 layers. Contrary to our initial hypothesis, cross-linguistic similarity did not uniformly impact idiom representations across all layers. While early and middle layers showed no significant differences among idiom categories, higher layers (from Layer 8 onwards) revealed more nuanced processing. Specifically, significant differences between the control category and idioms with similar meaning (SM), as well as between idioms with similar lexical items (SL) and those with similar semantics (SM) were observed. Our analysis revealed that early layers provided general representations, while higher layers showed increased differentiation between literal and figurative meanings. This was evidenced by a general decrease in cosine similarities from Layer 5 onwards, with Layer 8 demonstrating the lowest cosine similarities across all categories. Interestingly, a trend suggests that mBERT performs slightly better with more literal hints. The order of cosine similarity for the categorizations was: idioms with a degree of formal similarity, control idioms, idioms with both formal and semantic similarity, and finally idioms with only semantic similarity. These findings indicate that mBERT‘s processing of idioms evolves significantly across its layers, with cross-linguistic might affect more significantly in higher layers where more abstract semantic processing likely occurs.</abstract>
    <identifier type="citekey">rastegar-ramezani-2024-greek</identifier>
    <location>
        <url>https://aclanthology.org/2024.clicit-1.87/</url>
    </location>
    <part>
        <date>2024-12</date>
        <extent unit="page">
            <start>805</start>
            <end>812</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T From ‘It‘s All Greek to Me’ to ‘Nur Bahnhof Verstehen’: An Investigation of mBERT‘s Cross-Linguistic Capabilities
%A Rastegar, Aria
%A Ramezani, Pegah
%Y Dell’Orletta, Felice
%Y Lenci, Alessandro
%Y Montemagni, Simonetta
%Y Sprugnoli, Rachele
%S Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024)
%D 2024
%8 December
%I CEUR Workshop Proceedings
%C Pisa, Italy
%@ 979-12-210-7060-6
%F rastegar-ramezani-2024-greek
%X This study investigates the impact of cross-linguistic similarities on idiom representation in mBERT, focusing on English and German idioms categorized by different degrees of similarity. We aim to determine whether different degrees of cross-linguistic similarities significantly affect mBERT‘s representations and to observe how these representations change across its 12 layers. Contrary to our initial hypothesis, cross-linguistic similarity did not uniformly impact idiom representations across all layers. While early and middle layers showed no significant differences among idiom categories, higher layers (from Layer 8 onwards) revealed more nuanced processing. Specifically, significant differences between the control category and idioms with similar meaning (SM), as well as between idioms with similar lexical items (SL) and those with similar semantics (SM) were observed. Our analysis revealed that early layers provided general representations, while higher layers showed increased differentiation between literal and figurative meanings. This was evidenced by a general decrease in cosine similarities from Layer 5 onwards, with Layer 8 demonstrating the lowest cosine similarities across all categories. Interestingly, a trend suggests that mBERT performs slightly better with more literal hints. The order of cosine similarity for the categorizations was: idioms with a degree of formal similarity, control idioms, idioms with both formal and semantic similarity, and finally idioms with only semantic similarity. These findings indicate that mBERT‘s processing of idioms evolves significantly across its layers, with cross-linguistic might affect more significantly in higher layers where more abstract semantic processing likely occurs.
%U https://aclanthology.org/2024.clicit-1.87/
%P 805-812

Markdown (Informal)

[From ‘It’s All Greek to Me’ to ‘Nur Bahnhof Verstehen’: An Investigation of mBERT’s Cross-Linguistic Capabilities](https://aclanthology.org/2024.clicit-1.87/) (Rastegar & Ramezani, CLiC-it 2024)

• From ‘It’s All Greek to Me’ to ‘Nur Bahnhof Verstehen’: An Investigation of mBERT’s Cross-Linguistic Capabilities (Rastegar & Ramezani, CLiC-it 2024)

ACL

• Aria Rastegar and Pegah Ramezani. 2024. From ‘It’s All Greek to Me’ to ‘Nur Bahnhof Verstehen’: An Investigation of mBERT’s Cross-Linguistic Capabilities. In Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024), pages 805–812, Pisa, Italy. CEUR Workshop Proceedings.