Multilingual neural machine translation (MNMT) learns to translate multiple language pairs with a single model, potentially improving both the accuracy and the memory-efficiency of deployed models. However, the heavy data imbalance between languages
hinders the model from performing uniformly across language pairs. In this paper, we propose a new learning objective for MNMT based on distributionally robust optimization, which minimizes the worst-case expected loss over the set of language pairs. We further show how to practically optimize this objective for large translation corpora using an iterated best response scheme, which is both effective and incurs negligible additional computational cost compared to standard empirical risk minimization. We perform extensive experiments on three sets of languages from two datasets and show that our method consistently outperforms strong baseline methods in terms of average and per-language performance under both many-to-one and one-to-many translation settings.
We present the results of the first task on Large-Scale Multilingual Machine Translation. The task consists on the many-to-many evaluation of a single model across a variety of source and target languages. This year, the task consisted on three diffe
rent settings: (i) SMALL-TASK1 (Central/South-Eastern European Languages), (ii) the SMALL-TASK2 (South-East Asian Languages), and (iii) FULL-TASK (all 101 x 100 language pairs). All the tasks used the FLORES-101 dataset as the evaluation benchmark. To ensure the longevity of the dataset, the test sets were not publicly released and the models were evaluated in a controlled environment on Dynabench. There were a total of 10 participating teams for the tasks, with a total of 151 intermediate model submissions and 13 final models. This year's result show a significant improvement over the known base-lines with +17.8 BLEU for SMALL-TASK2, +10.6 for FULL-TASK and +3.6 for SMALL-TASK1.
Transformers that are pre-trained on multilingual corpora, such as, mBERT and XLM-RoBERTa, have achieved impressive cross-lingual transfer capabilities. In the zero-shot transfer setting, only English training data is used, and the fine-tuned model i
s evaluated on another target language. While this works surprisingly well, substantial variance has been observed in target language performance between different fine-tuning runs, and in the zero-shot setup, no target-language development data is available to select among multiple fine-tuned models. Prior work has relied on English dev data to select among models that are fine-tuned with different learning rates, number of steps and other hyperparameters, often resulting in suboptimal choices. In this paper, we show that it is possible to select consistently better models when small amounts of annotated data are available in auxiliary pivot languages. We propose a machine learning approach to model selection that uses the fine-tuned model's own internal representations to predict its cross-lingual capabilities. In extensive experiments we find that this method consistently selects better models than English validation data across twenty five languages (including eight low-resource languages), and often achieves results that are comparable to model selection using target language development data.
This paper describes the systems submitted to WAT 2021 MultiIndicMT shared task by IITP-MT team. We submit two multilingual Neural Machine Translation (NMT) systems (Indic-to-English and English-to-Indic). We romanize all Indic data and create subwor
d vocabulary which is shared between all Indic languages. We use back-translation approach to generate synthetic data which is appended to parallel corpus and used to train our models. The models are evaluated using BLEU, RIBES and AMFM scores with Indic-to-English model achieving 40.08 BLEU for Hindi-English pair and English-to-Indic model achieving 34.48 BLEU for English-Hindi pair. However, we observe that the shared romanized subword vocabulary is not helping English-to-Indic model at the time of generation, leading it to produce poor quality translations for Tamil, Telugu and Malayalam to English pairs with BLEU score of 8.51, 6.25 and 3.79 respectively.
The choice of parameter sharing strategy in multilingual machine translation models determines how optimally parameter space is used and hence, directly influences ultimate translation quality. Inspired by linguistic trees that show the degree of rel
atedness between different languages, the new general approach to parameter sharing in multilingual machine translation was suggested recently. The main idea is to use these expert language hierarchies as a basis for multilingual architecture: the closer two languages are, the more parameters they share. In this work, we test this idea using the Transformer architecture and show that despite the success in previous work there are problems inherent to training such hierarchical models. We demonstrate that in case of carefully chosen training strategy the hierarchical architecture can outperform bilingual models and multilingual models with full parameter sharing.
