On the Language Neutrality of Pre-trained Multilingual Representations

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F20%3A10424472" target="_blank" >RIV/00216208:11320/20:10424472 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.aclweb.org/anthology/2020.findings-emnlp.150/" target="_blank" >https://www.aclweb.org/anthology/2020.findings-emnlp.150/</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.18653/v1/2020.findings-emnlp.150" target="_blank" >10.18653/v1/2020.findings-emnlp.150</a>

Jazyk výsledku

angličtina

Název v původním jazyce

On the Language Neutrality of Pre-trained Multilingual Representations

Popis výsledku v původním jazyce

Multilingual contextual embeddings, such as multilingual BERT (mBERT) and XLM-RoBERTa, have proved useful for many multi-lingual tasks. Previous work probed the cross-linguality of the representations indirectly using zero-shot transfer learning on morphological and syntactic tasks. We instead focus on the language-neutrality of mBERT with respect to lexical semantics. Our results show that contextual embeddings are more language-neutral and in general more informative than aligned static word-type embeddings which are explicitly trained for language neutrality. Contextual embeddings are still by default only moderately language-neutral, however, we show two simple methods for achieving stronger language neutrality: first, by unsupervised centering of the representation for languages, and second by fitting an explicit projection on small parallel data. In addition, we show how to reach state-of-the-art accuracy on language identification and word alignment in parallel sentences.

Název v anglickém jazyce

On the Language Neutrality of Pre-trained Multilingual Representations

Popis výsledku anglicky

Multilingual contextual embeddings, such as multilingual BERT (mBERT) and XLM-RoBERTa, have proved useful for many multi-lingual tasks. Previous work probed the cross-linguality of the representations indirectly using zero-shot transfer learning on morphological and syntactic tasks. We instead focus on the language-neutrality of mBERT with respect to lexical semantics. Our results show that contextual embeddings are more language-neutral and in general more informative than aligned static word-type embeddings which are explicitly trained for language neutrality. Contextual embeddings are still by default only moderately language-neutral, however, we show two simple methods for achieving stronger language neutrality: first, by unsupervised centering of the representation for languages, and second by fitting an explicit projection on small parallel data. In addition, we show how to reach state-of-the-art accuracy on language identification and word alignment in parallel sentences.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/GA18-02196S" target="_blank" >GA18-02196S: Reprezentace lingvistické struktury v neuronových sítích</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2020

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Findings of the Association for Computational Linguistics: EMNLP 2020

ISBN

978-1-952148-90-3

ISSN

—

e-ISSN

—

Počet stran výsledku

12

Strana od-do

1663-1674

Název nakladatele

Association for Computational Linguistics

Místo vydání

Stroudsburg, PA, USA

Místo konání akce

Online

Datum konání akce

16. 11. 2020

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—