In the perspective of a layer normalization (LN) position, the architecture of Transformers can be categorized into two types: Post-LN and Pre-LN.Recent Transformers prefer to select Pre-LN because the training in Post-LN with deep Transformers, e.g., ten or more layers, often becomes unstable, resulting in useless models. However, in contrast, Post-LN has also consistently achieved better performance than Pre-LN in relatively shallow Transformers, e.g., six or fewer layers. This study first investigates the reason for these discrepant observations empirically and theoretically and discovers 1, the LN in Post-LN is the source of the vanishing gradient problem that mainly leads the unstable training whereas Pre-LN prevents it, and 2, Post-LN tends to preserve larger gradient norms in higher layers during the back-propagation that may lead an effective training. Exploiting the new findings, we propose a method that can equip both higher stability and effective training by a simple modification from Post-LN.We conduct experiments on a wide range of text generation tasks and demonstrate that our method outperforms Pre-LN, and stable training regardless of the shallow or deep layer settings.

Pretrained language models require the use of consistent segmentation (e.g., subword- or character-level segmentation) in pretraining and finetuning. In NLP, many tasks are modeled by subword-level segmentation better than by character-level segmentation. However, because of their format, several tasks require the use of character-level segmentation. Thus, in order to tackle both types of NLP tasks, language models must be independently pretrained for both subword and character-level segmentation. However, this is an inefficient and costly procedure. Instead, this paper proposes a method for training a language model with unified segmentation. This means that the trained model can be finetuned on both subword- and character-level segmentation. The principle of the method is to apply the subword regularization technique to generate a mixture of subword- and character-level segmentation. Through experiment on BERT models, we demonstrate that our method can halve the computational cost of pretraining.

This paper describes the NTT-Tohoku-TokyoTech-RIKEN (NT5) team’s submission system for the WMT’22 general translation task. This year, we focused on the English-to-Japanese and Japanese-to-English translation tracks. Our submission system consists of an ensemble of Transformer models with several extensions. We also applied data augmentation and selection techniques to obtain potentially effective training data for training individual Transformer models in the pre-training and fine-tuning scheme. Additionally, we report our trial of incorporating a reranking module and the reevaluated results of several techniques that have been recently developed and published.

Ensembling is a popular method used to improve performance as a last resort. However, ensembling multiple models finetuned from a single pretrained model has been not very effective; this could be due to the lack of diversity among ensemble members. This paper proposes Multi-Ticket Ensemble, which finetunes different subnetworks of a single pretrained model and ensembles them. We empirically demonstrated that winning-ticket subnetworks produced more diverse predictions than dense networks and their ensemble outperformed the standard ensemble in some tasks when accurate lottery tickets are found on the tasks.

We often use perturbations to regularize neural models. For neural encoder-decoders, previous studies applied the scheduled sampling (Bengio et al., 2015) and adversarial perturbations (Sato et al., 2019) as perturbations but these methods require considerable computational time. Thus, this study addresses the question of whether these approaches are efficient enough for training time. We compare several perturbations in sequence-to-sequence problems with respect to computational time. Experimental results show that the simple techniques such as word dropout (Gal and Ghahramani, 2016) and random replacement of input tokens achieve comparable (or better) scores to the recently proposed perturbations, even though these simple methods are faster.

Position representation is crucial for building position-aware representations in Transformers. Existing position representations suffer from a lack of generalization to test data with unseen lengths or high computational cost. We investigate shifted absolute position embedding (SHAPE) to address both issues. The basic idea of SHAPE is to achieve shift invariance, which is a key property of recent successful position representations, by randomly shifting absolute positions during training. We demonstrate that SHAPE is empirically comparable to its counterpart while being simpler and faster.

Masked language models (MLMs) have contributed to drastic performance improvements with regard to zero anaphora resolution (ZAR). To further improve this approach, in this study, we made two proposals. The first is a new pretraining task that trains MLMs on anaphoric relations with explicit supervision, and the second proposal is a new finetuning method that remedies a notorious issue, the pretrain-finetune discrepancy. Our experiments on Japanese ZAR demonstrated that our two proposals boost the state-of-the-art performance, and our detailed analysis provides new insights on the remaining challenges.

One critical issue of zero anaphora resolution (ZAR) is the scarcity of labeled data. This study explores how effectively this problem can be alleviated by data augmentation. We adopt a state-of-the-art data augmentation method, called the contextual data augmentation (CDA), that generates labeled training instances using a pretrained language model. The CDA has been reported to work well for several other natural language processing tasks, including text classification and machine translation. This study addresses two underexplored issues on CDA, that is, how to reduce the computational cost of data augmentation and how to ensure the quality of the generated data. We also propose two methods to adapt CDA to ZAR: [MASK]-based augmentation and linguistically-controlled masking. Consequently, the experimental results on Japanese ZAR show that our methods contribute to both the accuracy gainand the computation cost reduction. Our closer analysis reveals that the proposed method can improve the quality of the augmented training data when compared to the conventional CDA.

In this paper, we describe the submission of Tohoku-AIP-NTT to the WMT’20 news translation task. We participated in this task in two language pairs and four language directions: English <–> German and English <–> Japanese. Our system consists of techniques such as back-translation and fine-tuning, which are already widely adopted in translation tasks. We attempted to develop new methods for both synthetic data filtering and reranking. However, the methods turned out to be ineffective, and they provided us with no significant improvement over the baseline. We analyze these negative results to provide insights for future studies.

This paper investigates how to effectively incorporate a pre-trained masked language model (MLM), such as BERT, into an encoder-decoder (EncDec) model for grammatical error correction (GEC). The answer to this question is not as straightforward as one might expect because the previous common methods for incorporating a MLM into an EncDec model have potential drawbacks when applied to GEC. For example, the distribution of the inputs to a GEC model can be considerably different (erroneous, clumsy, etc.) from that of the corpora used for pre-training MLMs; however, this issue is not addressed in the previous methods. Our experiments show that our proposed method, where we first fine-tune a MLM with a given GEC corpus and then use the output of the fine-tuned MLM as additional features in the GEC model, maximizes the benefit of the MLM. The best-performing model achieves state-of-the-art performances on the BEA-2019 and CoNLL-2014 benchmarks. Our code is publicly available at: https://github.com/kanekomasahiro/bert-gec.

We present ESPnet-ST, which is designed for the quick development of speech-to-speech translation systems in a single framework. ESPnet-ST is a new project inside end-to-end speech processing toolkit, ESPnet, which integrates or newly implements automatic speech recognition, machine translation, and text-to-speech functions for speech translation. We provide all-in-one recipes including data pre-processing, feature extraction, training, and decoding pipelines for a wide range of benchmark datasets. Our reproducible results can match or even outperform the current state-of-the-art performances; these pre-trained models are downloadable. The toolkit is publicly available at https://github.com/espnet/espnet.

Existing approaches for grammatical error correction (GEC) largely rely on supervised learning with manually created GEC datasets. However, there has been little focus on verifying and ensuring the quality of the datasets, and on how lower-quality data might affect GEC performance. We indeed found that there is a non-negligible amount of “noise” where errors were inappropriately edited or left uncorrected. To address this, we designed a self-refinement method where the key idea is to denoise these datasets by leveraging the prediction consistency of existing models, and outperformed strong denoising baseline methods. We further applied task-specific techniques and achieved state-of-the-art performance on the CoNLL-2014, JFLEG, and BEA-2019 benchmarks. We then analyzed the effect of the proposed denoising method, and found that our approach leads to improved coverage of corrections and facilitated fluency edits which are reflected in higher recall and overall performance.

A regularization technique based on adversarial perturbation, which was initially developed in the field of image processing, has been successfully applied to text classification tasks and has yielded attractive improvements. We aim to further leverage this promising methodology into more sophisticated and critical neural models in the natural language processing field, i.e., neural machine translation (NMT) models. However, it is not trivial to apply this methodology to such models. Thus, this paper investigates the effectiveness of several possible configurations of applying the adversarial perturbation and reveals that the adversarial regularization technique can significantly and consistently improve the performance of widely used NMT models, such as LSTM-based and Transformer-based models.

The incorporation of pseudo data in the training of grammatical error correction models has been one of the main factors in improving the performance of such models. However, consensus is lacking on experimental configurations, namely, choosing how the pseudo data should be generated or used. In this study, these choices are investigated through extensive experiments, and state-of-the-art performance is achieved on the CoNLL-2014 test set (F0.5=65.0) and the official test set of the BEA-2019 shared task (F0.5=70.2) without making any modifications to the model architecture.

This paper describes the ESPnet submissions to the How2 Speech Translation task at IWSLT2019. In this year, we mainly build our systems based on Transformer architectures in all tasks and focus on the end-to-end speech translation (E2E-ST). We first compare RNN-based models and Transformer, and then confirm Transformer models significantly and consistently outperform RNN models in all tasks and corpora. Next, we investigate pre-training of E2E-ST models with the ASR and MT tasks. On top of the pre-training, we further explore knowledge distillation from the NMT model and the deeper speech encoder, and confirm drastic improvements over the baseline model. All of our codes are publicly available in ESPnet.

Developing a method for understanding the inner workings of black-box neural methods is an important research endeavor. Conventionally, many studies have used an attention matrix to interpret how Encoder-Decoder-based models translate a given source sentence to the corresponding target sentence. However, recent studies have empirically revealed that an attention matrix is not optimal for token-wise translation analyses. We propose a method that explicitly models the token-wise alignment between the source and target sequences to provide a better analysis. Experiments show that our method can acquire token-wise alignments that are superior to those of an attention mechanism.