Další formáty:
BibTeX
LaTeX
RIS
@inproceedings{1809743,
author = {Štefánik, Michal and Sojka, Petr},
address = {Brno},
booktitle = {Recent Advances in Slavonic Natural Language Processing (RASLAN 2021)},
editor = {Horák, Rychlý, Rambousek},
keywords = {Generalization; Debiasing; Domain extrapolation; Domain adaptation; Domain robustness; Neural language models},
howpublished = {tištěná verze "print"},
language = {eng},
location = {Brno},
isbn = {978-80-263-1670-1},
pages = {91-103},
publisher = {Tribun EU},
title = {Towards Domain Robustness of Neural Language Models},
url = {https://nlp.fi.muni.cz/raslan/raslan21.pdf#page=99},
year = {2021}
}
TY - JOUR
ID - 1809743
AU - Štefánik, Michal - Sojka, Petr
PY - 2021
TI - Towards Domain Robustness of Neural Language Models
PB - Tribun EU
CY - Brno
SN - 9788026316701
KW - Generalization
KW - Debiasing
KW - Domain extrapolation
KW - Domain adaptation
KW - Domain robustness
KW - Neural language models
UR - https://nlp.fi.muni.cz/raslan/raslan21.pdf#page=99
N2 - This work summarises recent progress in generalization evaluation and training of deep neural networks, categorized in data-centric and model-centric overviews. Grounded in the results of the referenced work, we propose three future directions towards reaching higher robustness of language models to an unknown domain or its adaptation to an existing domain of interest. In the example propositions that practically complement each of the directions, we introduce novel ideas of a) dynamic objective selection, b) language modeling respecting the token similarities to the ground truth and c) a framework of additive component of the loss utilizing the well-performing generalization measures.
ER -
ŠTEFÁNIK, Michal a Petr SOJKA. Towards Domain Robustness of Neural Language Models. In Horák, Rychlý, Rambousek. \textit{Recent Advances in Slavonic Natural Language Processing (RASLAN 2021)}. Brno: Tribun EU, 2021, s.~91-103. ISBN~978-80-263-1670-1.
|
