Vše

Co hledáte?

Vše
Projekty
Výsledky výzkumu
Subjekty

Rychlé hledání

  • Projekty podpořené TA ČR
  • Významné projekty
  • Projekty s nejvyšší státní podporou
  • Aktuálně běžící projekty

Chytré vyhledávání

  • Takto najdu konkrétní +slovo
  • Takto z výsledků -slovo zcela vynechám
  • “Takto můžu najít celou frázi”

Emerging Cross-lingual Structure in Pretrained Language Models

Identifikátory výsledku

  • Kód výsledku v IS VaVaI

    <a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F20%3A10426998" target="_blank" >RIV/00216208:11320/20:10426998 - isvavai.cz</a>

  • Výsledek na webu

    <a href="https://www.aclweb.org/anthology/2020.acl-main.536" target="_blank" >https://www.aclweb.org/anthology/2020.acl-main.536</a>

  • DOI - Digital Object Identifier

Alternativní jazyky

  • Jazyk výsledku

    angličtina

  • Název v původním jazyce

    Emerging Cross-lingual Structure in Pretrained Language Models

  • Popis výsledku v původním jazyce

    We study the problem of multilingual masked language modeling, i.e. the training of a single model on concatenated text from multiple languages, and present a detailed study of several factors that influence why these models are so effective for cross-lingual transfer. We show, contrary to what was previously hypothesized, that transfer is possible even when there is no shared vocabulary across the monolingual corpora and also when the text comes from very different domains. The only requirement is that there are some shared parameters in the top layers of the multi-lingual encoder. To better understand this result, we also show that representations from monolingual BERT models in different languages can be aligned post-hoc quite effectively, strongly suggesting that, much like for non-contextual word embeddings, there are universal latent symmetries in the learned embedding spaces. For multilingual masked language modeling, these symmetries are automatically discovered and aligned during the joint training process.

  • Název v anglickém jazyce

    Emerging Cross-lingual Structure in Pretrained Language Models

  • Popis výsledku anglicky

    We study the problem of multilingual masked language modeling, i.e. the training of a single model on concatenated text from multiple languages, and present a detailed study of several factors that influence why these models are so effective for cross-lingual transfer. We show, contrary to what was previously hypothesized, that transfer is possible even when there is no shared vocabulary across the monolingual corpora and also when the text comes from very different domains. The only requirement is that there are some shared parameters in the top layers of the multi-lingual encoder. To better understand this result, we also show that representations from monolingual BERT models in different languages can be aligned post-hoc quite effectively, strongly suggesting that, much like for non-contextual word embeddings, there are universal latent symmetries in the learned embedding spaces. For multilingual masked language modeling, these symmetries are automatically discovered and aligned during the joint training process.

Klasifikace

  • Druh

    O - Ostatní výsledky

  • CEP obor

  • OECD FORD obor

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

Návaznosti výsledku

  • Projekt

  • Návaznosti

Ostatní

  • Rok uplatnění

    2020

  • Kód důvěrnosti údajů

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