Low-resource Neural Machine Translation Research Articles

Low resource neural machine translation model optimization based on semantic confidence weighted alignment

Low resource neural machine translation model optimization based on semantic confidence weighted alignment

A Study for Enhancing Low-resource Thai-Myanmar-English Neural Machine Translation

Several methodologies have recently been proposed to enhance the performance of low-resource Neural Machine Translation (NMT). However, these techniques have yet to be explored thoroughly in the low-resource Thai and Myanmar languages. Therefore, we first applied augmentation techniques such as SwitchOut and Ciphertext Based Data Augmentation (CipherDAug) to improve NMT performance in these languages. Second, we enhanced the NMT performance by fine-tuning the pre-trained Multilingual Denoising BART model (mBART), where BART denotes Bidirectional and Auto-Regressive Transformer. We implemented three NMT systems: namely, Transformer+SwitchOut, Multi-Source Transformer+CipherDAug, and fine-tuned mBART in the bidirectional translations of Thai-English-Myanmar language pairs from the ASEAN-MT corpus. Experimental results showed that Multi-Source Transformer+CipherDAug significantly improved Bilingual Evaluation Understudy (BLEU), Character n-gram F-score (ChrF) , and Translation Error Rate (TER) scores over the first baseline Transformer and second baseline Edit-Based Transformer. The model achieved notable BLEU scores: 37.9 (English-to-Thai), 42.7 (Thai-to-English), 28.9 (English-to-Myanmar), 31.2 (Myanmar-to-English), 25.3 (Thai-to-Myanmar), and 25.5 (Myanmar-to-Thai). The fine-tuned mBART model also considerably outperformed the two baselines, except for the Myanmar-to-English pair. SwitchOut improved over the second baseline in all pairs and performed similarly to the first baseline in most cases. Last, we performed detailed analyses verifying that the CipherDAug and mBART models potentially facilitate improving low-resource NMT performance in Thai and Myanmar languages.

Based on Gated Dynamic Encoding Optimization, the LGE-Transformer Method for Low-Resource Neural Machine Translation

Based on Gated Dynamic Encoding Optimization, the LGE-Transformer Method for Low-Resource Neural Machine Translation

Filtered data augmentation approach based on model competence evaluation

The scale of parallel corpus plays an important role in training high-quality neural machine translation models. In order to expand the scale of parallel corpus in low-resource scenarios, researchers have proposed a series of data augmentation approaches, in which the most representative work is the back-translation. The back-translation approach uses a basic translation model to translate the target monolingual corpus into the source language, and then combines it into a pseudo parallel corpus to expand the training data. Due to the simple and efficient data expansion strategies, the back-translation approach becomes the mainstream data augmentation approach of neural machine translation. However, in low-resource scenarios, only low-precision basic translation models can be used. The pseudo-parallel corpus translated by this model contains low-quality sentence pairs, which inevitably introduces noises and leads to negative impacts to translation procedure. To improve the performance of low-resource neural machine translation, we propose a filtered data augmentation method based on the model competence evaluation, which can effectively improve the quality of training data by using both the filtered back-translation and dynamic evaluation of translation model capabilities. Comparative experiments show that the proposed approach can improve the quality of training data and the performance of low-resource neural machine translation.

Neural Machine Translation of Electrical Engineering with Fusion of Memory Information

This paper proposes a new neural machine translation model of electrical engineering that combines a transformer with gated recurrent unit (GRU) networks. By fusing global information and memory information, the model effectively improves the performance of low-resource neural machine translation. Unlike traditional transformers, our proposed model includes two different encoders: one is the global information encoder, which focuses on contextual information, and the other is the memory encoder, which is responsible for capturing recurrent memory information. The model with these two types of attention can encode both global and memory information and learn richer semantic knowledge. Because transformers require global attention calculation for each word position, the time and space complexity are both squared with the length of the source language sequence. When the length of the source language sequence becomes too long, the performance of the transformer will sharply decline. Therefore, we propose a memory information encoder based on the GRU to improve this drawback. The model proposed in this paper has a maximum improvement of 2.04 BLEU points over the baseline model in the field of electrical engineering with low resources.

Towards better Chinese-centric neural machine translation for low-resource languages

The last decade has witnessed enormous improvements in science and technology, stimulating the growing demand for economic and cultural exchanges in various countries. Building a neural machine translation (NMT) system has become an urgent trend, especially in the low-resource setting. However, recent work tends to study NMT systems for low-resource languages centered on English, while a few studies focus on low-resource NMT systems centered on other languages such as Chinese. To achieve this, the low-resource multilingual translation challenge of the 2021 iFLYTEK AI Developer Competition provides the Chinese-centric multilingual low-resource NMT tasks, where participants are required to build NMT systems based on the provided low-resource samples. In this paper, we present the competition-winning system that leverages monolingual word embeddings data enhancement, bilingual curriculum learning, and contrastive re-ranking. In addition, a new Incomplete-Trust (In-trust) loss function is proposed to replace the traditional cross-entropy loss when training. The extensive experimental results demonstrate that the implementation of these ideas leads to better performance than other state-of-the-art methods. All the experimental codes are released at: https://github.com/WENGSYX/Low-resource-text-translation.

Low-resource Multilingual Neural Translation Using Linguistic Feature-based Relevance Mechanisms

This article investigates approaches to effectively harness source-side linguistic features for low-resource multilingual neural machine translation (MNMT). Previous works focus on using various features of a word such as lemma, part-of-speech tag, dependency label, and so on, to improve translation quality in a low-resource scenario. However, these studies deal with bilingual translation and do not focus on using features in multilingual training setups. Our work focuses on this particular point and experiments with low-resource multilingual models incorporating source-side linguistic features. Although techniques for integrating features into an NMT model such as concatenation and feature relevance perform quite well in bilingual settings, they do not work well in multilingual settings. To remedy this, we propose the use of dummy features and language indicator features in MNMT models. Experiments are conducted on English to Asian language translation on a multilingual, multi-parallel corpus spanning English and eight Asian languages where for each language pair, the training data size does not exceed 20,000 parallel sentences. After establishing strong bilingual baselines using feature relevance mechanisms and multilingual baselines without any features, we show that our proposed dummy features and language indicator features, in combination with feature relevance mechanisms, yield significant improvements in BLEU points for all language pairs. We then analyze our models from the perspectives of model sizes, the impact of individual linguistic features, validation perplexity computed during training, visualization of the activations of the relevance mechanisms, and exhaustive tuning of hyperparameters. We also report preliminary results for multilingual multi-way models using linguistic features.

STA: An efficient data augmentation method for low-resource neural machine translation

Transformer-based neural machine translation (NMT) has achieved state-of-the-art performance in the NMT paradigm. However, it relies on the availability of copious parallel corpora. For low-resource language pairs, the amount of parallel data is insufficient, resulting in poor translation quality. To alleviate this issue, this paper proposes an efficient data augmentation (DA) method named STA. Firstly, the pseudo-parallel sentence pairs are generated by translating sentence trunks with the target-to-source NMT model. Furthermore, two strategies are introduced to merge the original data and pseudo-parallel corpus to augment the training set. Experimental results on simulated and real low-resource translation tasks show that the proposed method improves the translation quality over the strong baseline, and also outperforms other data augmentation methods. Moreover, the STA method can further improve the translation quality when combined with the back-translation method with the extra monolingual data.

Exploring Graph-based Transformer Encoder for Low-Resource Neural Machine Translation

The Transformer is commonly used in Neural Machine Translation (NMT), but it faces issues with over-parameterization in low-resource settings. This means that simply increasing the model parameters significantly will not lead to improved performance. In this study, we propose a graph-based approach that slightly increases the parameters while significantly outperforming the scaled version of the Transformer. We accomplish this by utilizing Graph Neural Networks to encode Universal Conceptual Cognitive Annotation (UCCA), allowing the linguistic features of UCCA to be incorporated into the word embeddings. This improves the performance of the NMT system since the word embedding is now more capable and informative. Experimental results demonstrate that the proposed method outperforms the scaled Transformer model by +0.4, +0.41, and +0.33 BLEU, respectively, in English-Vietnamese/French/Czech datasets. Furthermore, this method reduces the number of parameters by 47% when compared to the scaled Transformer. A thorough analysis of error patterns reveals that the proposed method provides structural awareness to translation systems. Our code is available at: https://github.com/nqbinh17/UCCA_GNN.

Neural Machine Translation for Low-resource Languages: A Survey

Neural Machine Translation (NMT) has seen tremendous growth in the last ten years since the early 2000s and has already entered a mature phase. While considered the most widely used solution for Machine Translation, its performance on low-resource language pairs remains sub-optimal compared to the high-resource counterparts due to the unavailability of large parallel corpora. Therefore, the implementation of NMT techniques for low-resource language pairs has been receiving the spotlight recently, thus leading to substantial research on this topic. This article presents a detailed survey of research advancements in low-resource language NMT (LRL-NMT) and quantitative analysis to identify the most popular techniques. We provide guidelines to select the possible NMT technique for a given LRL data setting based on our findings. We also present a holistic view of the LRL-NMT research landscape and provide recommendations to enhance the research efforts further.

Low-resource Neural Machine Translation: Methods and Trends

Neural Machine Translation (NMT) brings promising improvements in translation quality, but until recently, these models rely on large-scale parallel corpora. As such corpora only exist on a handful of language pairs, the translation performance is far from the desired effect in the majority of low-resource languages. Thus, developing low-resource language translation techniques is crucial and it has become a popular research field in neural machine translation. In this article, we make an overall review of existing deep learning techniques in low-resource NMT. We first show the research status as well as some widely used low-resource datasets. Then, we categorize the existing methods and show some representative works detailedly. Finally, we summarize the common characters among them and outline the future directions in this field.

Research on the Analysis of Correlation Factors of English Translation Ability Improvement Based on Deep Neural Network.

This paper adopts the algorithm of the deep neural network to conduct in-depth research and analysis on the factors associated with the improvement of English translation ability. This study focuses on text complexity, adding discourse complexity features in addition to focusing on lexical and syntactic dimensions, exploring the application of neural network algorithm in the construction of text complexity grading model based on feature optimization, and examining the performance and generalization ability of the model. The rationality of the grading of the material is verified. After determining the model input features and training corpus, different classification algorithms were used to build the models and compare their performance. Meanwhile, compared with the models constructed based on common traditional readability formulas and other single-dimensional features, the models constructed based on the feature set of this study have significant advantages, with 20 to 30 percentage points higher in each performance evaluation index. The pseudo-parallel corpus is constructed, back translation is performed after obtaining the pseudo-parallel corpus, and finally, the data migration effect is measured and recorded on the low-resource Chinese-English parallel corpus and Tibetan-Chinese parallel corpus, and the cycle continues until the model performance is no longer improved. The low-resource neural machine translation model based on model migration learning improved 3.97 and 2.64 BLEU values in the low-resource English translation task, respectively, and reduced the training time; based on this, the data migration learning method further improved 2.26 and 2.52 BLEU values.

An empirical study of low-resource neural machine translation of manipuri in multilingual settings

An empirical study of low-resource neural machine translation of manipuri in multilingual settings

Enhancing low-resource neural machine translation with syntax-graph guided self-attention

Most neural machine translation (NMT) models only rely on parallel sentence pairs, while the performance drops sharply in low-resource cases, as the models fail to mine the linguistry of the corpus. Incorporating prior monolingual knowledge explicitly, such as syntax, has been shown to be effective for NMT, particularly in low-resource scenarios. However, existing approaches have not exploited the full potential of the NMT architectures. In this paper, we present syntax-graph guided self-attention (SGSA): a neural network model that combines the source-side syntactic knowledge with multi-head self-attention. We introduce an additional syntax-aware localness modeling as a bias, which indicates that the syntactically relevant parts need to be paid more attention to. The bias is then incorporated into the original attention distribution to form a revised distribution. Moreover, to maintain the strength of capturing the meaningful semantic representations of source-sentence, we adopt a node random dropping strategy in multi-head self-attention subnetworks. Extensive experiments on several standard small-scale datasets demonstrate that SGSA can significantly improve the performance of Transformer-based NMT, and is also superior to the previous syntax-dependent state-of-the-art.

Improving Neural Machine Translation for Low Resource Algerian Dialect by Transductive Transfer Learning Strategy.

This study is the first work on a transductive transfer learning approach for low-resource neural machine translation applied to the Algerian Arabic dialect. The transductive approach is based on a fine-tuning transfer learning strategy that transfers knowledge from the parent model to the child model. This strategy helps to solve the learning problem using limited parallel corpora. We tested the approach on a sequence-to-sequence model with and without the Attention mechanism. We first trained the models on a parallel multi-dialects Arabic corpus and then switch them to a low-resource of the Algerian dialect. Transductive transfer learning raises the BLEU score for the Seq2Seq model from 0.3 to more than 34, and for the Attentional-Seq2Seq model from less than 17 to more than 35. The obtained results prove the validity of this approach.

Enriching the transfer learning with pre-trained lexicon embedding for low-resource neural machine translation

Most State-Of-The-Art (SOTA) Neural Machine Translation (NMT) systems today achieve outstanding results based only on large parallel corpora. The large-scale parallel corpora for high-resource languages is easily obtainable. However, the translation quality of NMT for morphologically rich languages is still unsatisfactory, mainly because of the data sparsity problem encountered in Low-Resource Languages (LRLs). In the low-resource NMT paradigm, Transfer Learning (TL) has been developed into one of the most efficient methods. It is difficult to train the model on high-resource languages to include the information in both parent and child models, as well as the initially trained model that only contains the lexicon features and word embeddings of the parent model instead of the child languages feature. In this work, we aim to address this issue by proposing the language-independent Hybrid Transfer Learning (HTL) method for LRLs by sharing lexicon embedding between parent and child languages without leveraging back translation or manually injecting noises. First, we train the High-Resource Languages (HRLs) as the parent model with its vocabularies. Then, we combine the parent and child language pairs using the oversampling method to train the hybrid model initialized by the previously parent model. Finally, we fine-tune the morphologically rich child model using a hybrid model. Besides, we explore some exciting discoveries on the original TL approach. Experimental results show that our model consistently outperforms five SOTA methods in two languages Azerbaijani (Az) and Uzbek (Uz). Meanwhile, our approach is practical and significantly better, achieving improvements of up to 4.94 and 4.84 BLEU points for low-resource child languages Az ! Zh and Uz ! Zh, respectively.

Linguistically Driven Multi-Task Pre-Training for Low-Resource Neural Machine Translation

In the present study, we propose novel sequence-to-sequence pre-training objectives for low-resource machine translation (NMT): Japanese-specific sequence to sequence (JASS) for language pairs involving Japanese as the source or target language, and English-specific sequence to sequence (ENSS) for language pairs involving English. JASS focuses on masking and reordering Japanese linguistic units known as bunsetsu, whereas ENSS is proposed based on phrase structure masking and reordering tasks. Experiments on ASPEC Japanese–English & Japanese–Chinese, Wikipedia Japanese–Chinese, News English–Korean corpora demonstrate that JASS and ENSS outperform MASS and other existing language-agnostic pre-training methods by up to +2.9 BLEU points for the Japanese–English tasks, up to +7.0 BLEU points for the Japanese–Chinese tasks and up to +1.3 BLEU points for English–Korean tasks. Empirical analysis, which focuses on the relationship between individual parts in JASS and ENSS, reveals the complementary nature of the subtasks of JASS and ENSS. Adequacy evaluation using LASER, human evaluation, and case studies reveals that our proposed methods significantly outperform pre-training methods without injected linguistic knowledge and they have a larger positive impact on the adequacy as compared to the fluency.

Research on Low Resource Neural Machine Translation Based on Non-autoregressive Model

The autoregressive model can’t make full use of context information because of its single direction of generation, and the autoregressive method can’t perform parallel computation in decoding, which affects the efficiency of translation generation. Therefore, we explore a non-autoregressive translation generation method based on insertion and deletion in low-resource languages, which decomposes translation generation into three steps: deletion-insertion-generation. Therefore, the dynamic editing of the translation can be realized in the iterative updating process. At the same time, each step can be calculated in parallel, which improves the decoding efficiency. In order to reduce the complexity of data sets in non-autoregressive model training, we have trained Uyghur-Chinese training data with sequence-level knowledge distillation. Experiments on Uyghur-Chinese, English-Romanian distilled data sets and standard data sets verify the effectiveness of the non-autoregressive method.

Augmenting training data with syntactic phrasal-segments in low-resource neural machine translation

Augmenting training data with syntactic phrasal-segments in low-resource neural machine translation

Recent advances of low-resource neural machine translation

Recent advances of low-resource neural machine translation