Шуды басатын тіл моделі
method
ID: denoisinglanguagemodel
Related Papers
MAAM: A Morphology-Aware Alignment Model for Unsupervised Bilingual Lexicon Induction
AbstractThe task of unsupervised bilingual lexicon induction (UBLI) aims to induce word translations from monolingual corpora in two languages. Previous work has shown that morphological variation is an intractable challenge …...
Data Synthesis and Iterative Refinement for Neural Semantic Parsing without Annotated Logical Forms
Abstract“Semantic parsing aims to convert natural language utterances to logical forms. A critical challenge for constructing semantic parsers is the lack of labeled data. In this paper, we propose a …...
Instance Regularization for Discriminative Language Model Pre-training
AbstractDiscriminative pre-trained language models (PrLMs) can be generalized as denoising auto-encoders that work with two procedures, ennoising and denoising. First, an ennoising process corrupts texts with arbitrary noising functions to …...