1 |
Alternated Training with Synthetic and Authentic Data for Neural Machine Translation ...
|
|
|
|
Abstract:
While synthetic bilingual corpora have demonstrated their effectiveness in low-resource neural machine translation (NMT), adding more synthetic data often deteriorates translation performance. In this work, we propose alternated training with synthetic and authentic data for NMT. The basic idea is to alternate synthetic and authentic corpora iteratively during training. Compared with previous work, we introduce authentic data as guidance to prevent the training of NMT models from being disturbed by noisy synthetic data. Experiments on Chinese-English and German-English translation tasks show that our approach improves the performance over several strong baselines. We visualize the BLEU landscape to further investigate the role of authentic and synthetic data during alternated training. From the visualization, we find that authentic data helps to direct the NMT model parameters towards points with higher BLEU scores and leads to consistent translation performance improvement. ... : ACL 2021, Short Findings ...
|
|
Keyword:
Computation and Language cs.CL; FOS Computer and information sciences
|
|
URL: https://arxiv.org/abs/2106.08582 https://dx.doi.org/10.48550/arxiv.2106.08582
|
|
BASE
|
|
Hide details
|
|
2 |
CPM-2: Large-scale Cost-effective Pre-trained Language Models ...
|
|
|
|
BASE
|
|
Show details
|
|
3 |
VISITRON: Visual Semantics-Aligned Interactively Trained Object-Navigator ...
|
|
|
|
BASE
|
|
Show details
|
|
4 |
Assessing Multilingual Fairness in Pre-trained Multimodal Representations ...
|
|
|
|
BASE
|
|
Show details
|
|
5 |
Dialog{S}um: {A} Real-Life Scenario Dialogue Summarization Dataset ...
|
|
|
|
BASE
|
|
Show details
|
|
6 |
Transfer Learning for Sequence Generation: from Single-source to Multi-source ...
|
|
|
|
BASE
|
|
Show details
|
|
8 |
Segment, Mask, and Predict: Augmenting Chinese Word Segmentation with Self-Supervision ...
|
|
|
|
BASE
|
|
Show details
|
|
9 |
Learning to Selectively Learn for Weakly-supervised Paraphrase Generation ...
|
|
|
|
BASE
|
|
Show details
|
|
10 |
SWSR: A Chinese Dataset and Lexicon for Online Sexism Detection ...
|
|
|
|
BASE
|
|
Show details
|
|
11 |
Analyzing the Limits of Self-Supervision in Handling Bias in Language ...
|
|
|
|
BASE
|
|
Show details
|
|
12 |
Statistically significant detection of semantic shifts using contextual word embeddings ...
|
|
|
|
BASE
|
|
Show details
|
|
13 |
SWSR: A Chinese Dataset and Lexicon for Online Sexism Detection ...
|
|
|
|
BASE
|
|
Show details
|
|
14 |
Statistically Significant Detection of Semantic Shifts using Contextual Word Embeddings ...
|
|
|
|
BASE
|
|
Show details
|
|
15 |
Leveraging Word-Formation Knowledge for Chinese Word Sense Disambiguation ...
|
|
|
|
BASE
|
|
Show details
|
|
16 |
SWSR: A Chinese Dataset and Lexicon for Online Sexism Detection ...
|
|
|
|
BASE
|
|
Show details
|
|
17 |
SWSR: A Chinese Dataset and Lexicon for Sexist Hate Speech Detection ...
|
|
|
|
BASE
|
|
Show details
|
|
18 |
SWSR: A Chinese Dataset and Lexicon for Sexist Hate Speech Detection ...
|
|
|
|
BASE
|
|
Show details
|
|
19 |
Tapping into non-English-language science for the conservation of global biodiversity. ...
|
|
|
|
BASE
|
|
Show details
|
|
20 |
Tapping into non-English-language science for the conservation of global biodiversity. ...
|
|
|
|
BASE
|
|
Show details
|
|
|
|