1 |
The FLORES-101 Evaluation Benchmark for Low-Resource and Multilingual Machine Translation ...
|
|
|
|
BASE
|
|
Show details
|
|
2 |
LAWDR: Language-Agnostic Weighted Document Representations from Pre-trained Models ...
|
|
|
|
BASE
|
|
Show details
|
|
3 |
Classification-based Quality Estimation: Small and Efficient Models for Real-world Applications ...
|
|
|
|
BASE
|
|
Show details
|
|
5 |
Findings of the AmericasNLP 2021 Shared Task on Open Machine Translation for Indigenous Languages of the Americas ...
|
|
|
|
BASE
|
|
Show details
|
|
6 |
Alternative Input Signals Ease Transfer in Multilingual Machine Translation ...
|
|
|
|
BASE
|
|
Show details
|
|
7 |
Adapting High-resource NMT Models to Translate Low-resource Related Languages without Parallel Data ...
|
|
|
|
BASE
|
|
Show details
|
|
8 |
Findings of the WMT 2021 Shared Task on Quality Estimation ...
|
|
|
|
BASE
|
|
Show details
|
|
9 |
AmericasNLI: Evaluating Zero-shot Natural Language Understanding of Pretrained Multilingual Models in Truly Low-resource Languages ...
|
|
Ebrahimi, Abteen; Mager, Manuel; Oncevay, Arturo; Chaudhary, Vishrav; Chiruzzo, Luis; Fan, Angela; Ortega, John; Ramos, Ricardo; Rios, Annette; Meza-Ruiz, Ivan; Giménez-Lugo, Gustavo A.; Mager, Elisabeth; Neubig, Graham; Palmer, Alexis; Coto-Solano, Rolando; Vu, Ngoc Thang; Kann, Katharina. - : arXiv, 2021
|
|
Abstract:
Pretrained multilingual models are able to perform cross-lingual transfer in a zero-shot setting, even for languages unseen during pretraining. However, prior work evaluating performance on unseen languages has largely been limited to low-level, syntactic tasks, and it remains unclear if zero-shot learning of high-level, semantic tasks is possible for unseen languages. To explore this question, we present AmericasNLI, an extension of XNLI (Conneau et al., 2018) to 10 indigenous languages of the Americas. We conduct experiments with XLM-R, testing multiple zero-shot and translation-based approaches. Additionally, we explore model adaptation via continued pretraining and provide an analysis of the dataset by considering hypothesis-only models. We find that XLM-R's zero-shot performance is poor for all 10 languages, with an average performance of 38.62%. Continued pretraining offers improvements, with an average accuracy of 44.05%. Surprisingly, training on poorly translated data by far outperforms all other ... : Accepted to ACL 2022 ...
|
|
Keyword:
Computation and Language cs.CL; FOS Computer and information sciences
|
|
URL: https://arxiv.org/abs/2104.08726 https://dx.doi.org/10.48550/arxiv.2104.08726
|
|
BASE
|
|
Hide details
|
|
10 |
Findings of the WMT 2021 shared task on quality estimation
|
|
|
|
In: 689 ; 730 (2021)
|
|
BASE
|
|
Show details
|
|
11 |
Multilingual Translation with Extensible Multilingual Pretraining and Finetuning ...
|
|
|
|
BASE
|
|
Show details
|
|
12 |
MLQE-PE: A Multilingual Quality Estimation and Post-Editing Dataset ...
|
|
|
|
BASE
|
|
Show details
|
|
13 |
Beyond English-Centric Multilingual Machine Translation ...
|
|
|
|
BASE
|
|
Show details
|
|
14 |
Unsupervised quality estimation for neural machine translation
|
|
|
|
In: 8 ; 539 ; 555 (2020)
|
|
BASE
|
|
Show details
|
|
15 |
An exploratory study on multilingual quality estimation
|
|
|
|
In: 366 ; 377 (2020)
|
|
BASE
|
|
Show details
|
|
16 |
BERGAMOT-LATTE submissions for the WMT20 quality estimation shared task
|
|
|
|
In: 1010 ; 1017 (2020)
|
|
BASE
|
|
Show details
|
|
17 |
Findings of the WMT 2020 shared task on quality estimation
|
|
|
|
In: 743 ; 764 (2020)
|
|
BASE
|
|
Show details
|
|
18 |
MLQE-PE: A multilingual quality estimation and post-editing dataset
|
|
|
|
BASE
|
|
Show details
|
|
19 |
Unsupervised Cross-lingual Representation Learning at Scale ...
|
|
|
|
BASE
|
|
Show details
|
|
20 |
WikiMatrix: Mining 135M Parallel Sentences in 1620 Language Pairs from Wikipedia ...
|
|
|
|
BASE
|
|
Show details
|
|
|
|