Can You Tell Me How to Get Past Sesame Street? Sentence-Level Pretraining Beyond Language Modeling

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61988987%3A17610%2F19%3AA20020EU" target="_blank" >RIV/61988987:17610/19:A20020EU - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.18653/v1/P19-1439" target="_blank" >http://dx.doi.org/10.18653/v1/P19-1439</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.18653/v1/P19-1439" target="_blank" >10.18653/v1/P19-1439</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Can You Tell Me How to Get Past Sesame Street? Sentence-Level Pretraining Beyond Language Modeling

Popis výsledku v původním jazyce

Natural language understanding has recently seen a surge of progress with the use of sentence encoders like ELMo (Peters et al., 2018a) and BERT (Devlin et al., 2019) which are pretrained on variants of language modeling. We conduct the first large-scale systematic study of candidate pretraining tasks, comparing 19 different tasks both as alternatives and complements to language modeling. Our primary results support the use language modeling, especially when combined with pretraining on additional labeled-data tasks. However, our results are mixed across pretraining tasks and show some concerning trends: In ELMo's pretrain-then-freeze paradigm, random baselines are worryingly strong and results vary strikingly across target tasks. In addition, fine-tuning BERT on an intermediate task often negatively impacts downstream transfer. In a more positive trend, we see modest gains from multitask training, suggesting the development of more sophisticated multitask and transfer learning techniques as an avenue for further research.

Název v anglickém jazyce

Can You Tell Me How to Get Past Sesame Street? Sentence-Level Pretraining Beyond Language Modeling

Popis výsledku anglicky

Natural language understanding has recently seen a surge of progress with the use of sentence encoders like ELMo (Peters et al., 2018a) and BERT (Devlin et al., 2019) which are pretrained on variants of language modeling. We conduct the first large-scale systematic study of candidate pretraining tasks, comparing 19 different tasks both as alternatives and complements to language modeling. Our primary results support the use language modeling, especially when combined with pretraining on additional labeled-data tasks. However, our results are mixed across pretraining tasks and show some concerning trends: In ELMo's pretrain-then-freeze paradigm, random baselines are worryingly strong and results vary strikingly across target tasks. In addition, fine-tuning BERT on an intermediate task often negatively impacts downstream transfer. In a more positive trend, we see modest gains from multitask training, suggesting the development of more sophisticated multitask and transfer learning techniques as an avenue for further research.

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

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2019

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of the 57th Conference of the Association for Computational Linguistics

ISBN

978-1-950737-48-2

ISSN

—

e-ISSN

—

Počet stran výsledku

12

Strana od-do

4465-4476

Název nakladatele

Association for Computational Linguistics 2019

Místo vydání

Florence

Místo konání akce

Florence

Datum konání akce

28. 7. 2019

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000493046106098