Xu Tan

Zixin Zeng, Rui Wang, Yichong Leng et al. · microsoft-research

While Neural Machine Translation(NMT) has achieved great progress in recent years, it still suffers from inaccurate translation of entities (e.g., person/organization name, location), due to the lack of entity training instances. When we humans encounter an unknown entity during translation, we usually first look up in a dictionary and then organize the entity translation together with the translations of other parts to form a smooth target sentence. Inspired by this translation process, we propose an Extract-and-Attend approach to enhance entity translation in NMT, where the translation candidates of source entities are first extracted from a dictionary and then attended to by the NMT model to generate the target sentence. Specifically, the translation candidates are extracted by first detecting the entities in a source sentence and then translating the entities through looking up in a dictionary. Then, the extracted candidates are added as a prefix of the decoder input to be attended to by the decoder when generating the target sentence through self-attention. Experiments conducted on En-Zh and En-Ru demonstrate that the proposed method is effective on improving both the translation accuracy of entities and the overall translation quality, with up to 35% reduction on entity error rate and 0.85 gain on BLEU and 13.8 gain on COMET.

Xu Tan, Tao Qin, Frank Soong et al.

Text to speech (TTS), or speech synthesis, which aims to synthesize intelligible and natural speech given text, is a hot research topic in speech, language, and machine learning communities and has broad applications in the industry. As the development of deep learning and artificial intelligence, neural network-based TTS has significantly improved the quality of synthesized speech in recent years. In this paper, we conduct a comprehensive survey on neural TTS, aiming to provide a good understanding of current research and future trends. We focus on the key components in neural TTS, including text analysis, acoustic models and vocoders, and several advanced topics, including fast TTS, low-resource TTS, robust TTS, expressive TTS, and adaptive TTS, etc. We further summarize resources related to TTS (e.g., datasets, opensource implementations) and discuss future research directions. This survey can serve both academic researchers and industry practitioners working on TTS.

Renqian Luo, Xu Tan, Rui Wang et al.

Neural architecture search (NAS) with an accuracy predictor that predicts the accuracy of candidate architectures has drawn increasing attention due to its simplicity and effectiveness. Previous works usually employ neural network-based predictors which require more delicate design and are easy to overfit. Considering that most architectures are represented as sequences of discrete symbols which are more like tabular data and preferred by non-neural predictors, in this paper, we study an alternative approach which uses non-neural model for accuracy prediction. Specifically, as decision tree based models can better handle tabular data, we leverage gradient boosting decision tree (GBDT) as the predictor for NAS. We demonstrate that the GBDT predictor can achieve comparable (if not better) prediction accuracy than neural network based predictors. Moreover, considering that a compact search space can ease the search process, we propose to prune the search space gradually according to important features derived from GBDT. In this way, NAS can be performed by first pruning the search space and then searching a neural architecture, which is more efficient and effective. Experiments on NASBench-101 and ImageNet demonstrate the effectiveness of using GBDT as predictor for NAS: (1) On NASBench-101, it is 22x, 8x, and 6x more sample efficient than random search, regularized evolution, and Monte Carlo Tree Search (MCTS) in finding the global optimum; (2) It achieves 24.2% top-1 error rate on ImageNet, and further achieves 23.4% top-1 error rate on ImageNet when enhanced with search space pruning. Code is provided at https://github.com/renqianluo/GBDT-NAS.

Jingxuan Wei, Linzhuang Sun, Xu Tan et al.

Knowledge distillation, a technique for model compression and performance enhancement, has gained significant traction in Neural Machine Translation (NMT). However, existing research primarily focuses on empirical applications, and there is a lack of comprehensive understanding of how student model capacity, data complexity, and decoding strategies collectively influence distillation effectiveness. Addressing this gap, our study conducts an in-depth investigation into these factors, particularly focusing on their interplay in word-level and sequence-level distillation within NMT. Through extensive experimentation across datasets like IWSLT13 En$\rightarrow$Fr, IWSLT14 En$\rightarrow$De, and others, we empirically validate hypotheses related to the impact of these factors on knowledge distillation. Our research not only elucidates the significant influence of model capacity, data complexity, and decoding strategies on distillation effectiveness but also introduces a novel, optimized distillation approach. This approach, when applied to the IWSLT14 de$\rightarrow$en translation task, achieves state-of-the-art performance, demonstrating its practical efficacy in advancing the field of NMT.

Bei Li, Yi Jing, Xu Tan et al.

Learning multiscale Transformer models has been evidenced as a viable approach to augmenting machine translation systems. Prior research has primarily focused on treating subwords as basic units in developing such systems. However, the incorporation of fine-grained character-level features into multiscale Transformer has not yet been explored. In this work, we present a \textbf{S}low-\textbf{F}ast two-stream learning model, referred to as Tran\textbf{SF}ormer, which utilizes a ``slow'' branch to deal with subword sequences and a ``fast'' branch to deal with longer character sequences. This model is efficient since the fast branch is very lightweight by reducing the model width, and yet provides useful fine-grained features for the slow branch. Our TranSFormer shows consistent BLEU improvements (larger than 1 BLEU point) on several machine translation benchmarks.

Jin Xu, Xu Tan, Kaitao Song et al.

Weight sharing is a popular approach to reduce the cost of neural architecture search (NAS) by reusing the weights of shared operators from previously trained child models. However, the rank correlation between the estimated accuracy and ground truth accuracy of those child models is low due to the interference among different child models caused by weight sharing. In this paper, we investigate the interference issue by sampling different child models and calculating the gradient similarity of shared operators, and observe: 1) the interference on a shared operator between two child models is positively correlated with the number of different operators; 2) the interference is smaller when the inputs and outputs of the shared operator are more similar. Inspired by these two observations, we propose two approaches to mitigate the interference: 1) MAGIC-T: rather than randomly sampling child models for optimization, we propose a gradual modification scheme by modifying one operator between adjacent optimization steps to minimize the interference on the shared operators; 2) MAGIC-A: forcing the inputs and outputs of the operator across all child models to be similar to reduce the interference. Experiments on a BERT search space verify that mitigating interference via each of our proposed methods improves the rank correlation of super-pet and combining both methods can achieve better results. Our discovered architecture outperforms RoBERTa$_{\rm base}$ by 1.1 and 0.6 points and ELECTRA$_{\rm base}$ by 1.6 and 1.1 points on the dev and test set of GLUE benchmark. Extensive results on the BERT compression, reading comprehension and ImageNet task demonstrate the effectiveness and generality of our proposed methods.

Rui Wang, Xu Tan, Renqian Luo et al.

Neural approaches have achieved state-of-the-art accuracy on machine translation but suffer from the high cost of collecting large scale parallel data. Thus, a lot of research has been conducted for neural machine translation (NMT) with very limited parallel data, i.e., the low-resource setting. In this paper, we provide a survey for low-resource NMT and classify related works into three categories according to the auxiliary data they used: (1) exploiting monolingual data of source and/or target languages, (2) exploiting data from auxiliary languages, and (3) exploiting multi-modal data. We hope that our survey can help researchers to better understand this field and inspire them to design better algorithms, and help industry practitioners to choose appropriate algorithms for their applications.

Yuzi Yan, Xu Tan, Bohan Li et al.

While recent text to speech (TTS) models perform very well in synthesizing reading-style (e.g., audiobook) speech, it is still challenging to synthesize spontaneous-style speech (e.g., podcast or conversation), mainly because of two reasons: 1) the lack of training data for spontaneous speech; 2) the difficulty in modeling the filled pauses (um and uh) and diverse rhythms in spontaneous speech. In this paper, we develop AdaSpeech 3, an adaptive TTS system that fine-tunes a well-trained reading-style TTS model for spontaneous-style speech. Specifically, 1) to insert filled pauses (FP) in the text sequence appropriately, we introduce an FP predictor to the TTS model; 2) to model the varying rhythms, we introduce a duration predictor based on mixture of experts (MoE), which contains three experts responsible for the generation of fast, medium and slow speech respectively, and fine-tune it as well as the pitch predictor for rhythm adaptation; 3) to adapt to other speaker timbre, we fine-tune some parameters in the decoder with few speech data. To address the challenge of lack of training data, we mine a spontaneous speech dataset to support our research this work and facilitate future research on spontaneous TTS. Experiments show that AdaSpeech 3 synthesizes speech with natural FP and rhythms in spontaneous styles, and achieves much better MOS and SMOS scores than previous adaptive TTS systems.

Lanqing Xue, Kaitao Song, Duocai Wu et al.

Rap generation, which aims to produce lyrics and corresponding singing beats, needs to model both rhymes and rhythms. Previous works for rap generation focused on rhyming lyrics but ignored rhythmic beats, which are important for rap performance. In this paper, we develop DeepRapper, a Transformer-based rap generation system that can model both rhymes and rhythms. Since there is no available rap dataset with rhythmic beats, we develop a data mining pipeline to collect a large-scale rap dataset, which includes a large number of rap songs with aligned lyrics and rhythmic beats. Second, we design a Transformer-based autoregressive language model which carefully models rhymes and rhythms. Specifically, we generate lyrics in the reverse order with rhyme representation and constraint for rhyme enhancement and insert a beat symbol into lyrics for rhythm/beat modeling. To our knowledge, DeepRapper is the first system to generate rap with both rhymes and rhythms. Both objective and subjective evaluations demonstrate that DeepRapper generates creative and high-quality raps with rhymes and rhythms. Code will be released on GitHub.

Jin Xu, Xu Tan, Renqian Luo et al.

While pre-trained language models (e.g., BERT) have achieved impressive results on different natural language processing tasks, they have large numbers of parameters and suffer from big computational and memory costs, which make them difficult for real-world deployment. Therefore, model compression is necessary to reduce the computation and memory cost of pre-trained models. In this work, we aim to compress BERT and address the following two challenging practical issues: (1) The compression algorithm should be able to output multiple compressed models with different sizes and latencies, in order to support devices with different memory and latency limitations; (2) The algorithm should be downstream task agnostic, so that the compressed models are generally applicable for different downstream tasks. We leverage techniques in neural architecture search (NAS) and propose NAS-BERT, an efficient method for BERT compression. NAS-BERT trains a big supernet on a search space containing a variety of architectures and outputs multiple compressed models with adaptive sizes and latency. Furthermore, the training of NAS-BERT is conducted on standard self-supervised pre-training tasks (e.g., masked language model) and does not depend on specific downstream tasks. Thus, the compressed models can be used across various downstream tasks. The technical challenge of NAS-BERT is that training a big supernet on the pre-training task is extremely costly. We employ several techniques including block-wise search, search space pruning, and performance approximation to improve search efficiency and accuracy. Extensive experiments on GLUE and SQuAD benchmark datasets demonstrate that NAS-BERT can find lightweight models with better accuracy than previous approaches, and can be directly applied to different downstream tasks with adaptive model sizes for different requirements of memory or latency.

Yuzi Yan, Xu Tan, Bohan Li et al.

Text to speech (TTS) is widely used to synthesize personal voice for a target speaker, where a well-trained source TTS model is fine-tuned with few paired adaptation data (speech and its transcripts) on this target speaker. However, in many scenarios, only untranscribed speech data is available for adaptation, which brings challenges to the previous TTS adaptation pipelines (e.g., AdaSpeech). In this paper, we develop AdaSpeech 2, an adaptive TTS system that only leverages untranscribed speech data for adaptation. Specifically, we introduce a mel-spectrogram encoder to a well-trained TTS model to conduct speech reconstruction, and at the same time constrain the output sequence of the mel-spectrogram encoder to be close to that of the original phoneme encoder. In adaptation, we use untranscribed speech data for speech reconstruction and only fine-tune the TTS decoder. AdaSpeech 2 has two advantages: 1) Pluggable: our system can be easily applied to existing trained TTS models without re-training. 2) Effective: our system achieves on-par voice quality with the transcribed TTS adaptation (e.g., AdaSpeech) with the same amount of untranscribed data, and achieves better voice quality than previous untranscribed adaptation methods. Synthesized speech samples can be found at https://speechresearch.github.io/adaspeech2/.

Yichong Leng, Xu Tan, Sheng Zhao et al.

Mean opinion score (MOS) is a popular subjective metric to assess the quality of synthesized speech, and usually involves multiple human judges to evaluate each speech utterance. To reduce the labor cost in MOS test, multiple methods have been proposed to automatically predict MOS scores. To our knowledge, for a speech utterance, all previous works only used the average of multiple scores from different judges as the training target and discarded the score of each individual judge, which did not well exploit the precious MOS training data. In this paper, we propose MBNet, a MOS predictor with a mean subnet and a bias subnet to better utilize every judge score in MOS datasets, where the mean subnet is used to predict the mean score of each utterance similar to that in previous works, and the bias subnet to predict the bias score (the difference between the mean score and each individual judge score) and capture the personal preference of individual judges. Experiments show that compared with MOSNet baseline that only leverages mean score for training, MBNet improves the system-level spearmans rank correlation co-efficient (SRCC) by 2.9% on VCC 2018 dataset and 6.7% on VCC 2016 dataset.

Renqian Luo, Xu Tan, Rui Wang et al.

Text to speech (TTS) has been broadly used to synthesize natural and intelligible speech in different scenarios. Deploying TTS in various end devices such as mobile phones or embedded devices requires extremely small memory usage and inference latency. While non-autoregressive TTS models such as FastSpeech have achieved significantly faster inference speed than autoregressive models, their model size and inference latency are still large for the deployment in resource constrained devices. In this paper, we propose LightSpeech, which leverages neural architecture search~(NAS) to automatically design more lightweight and efficient models based on FastSpeech. We first profile the components of current FastSpeech model and carefully design a novel search space containing various lightweight and potentially effective architectures. Then NAS is utilized to automatically discover well performing architectures within the search space. Experiments show that the model discovered by our method achieves 15x model compression ratio and 6.5x inference speedup on CPU with on par voice quality. Audio demos are provided at https://speechresearch.github.io/lightspeech.

Yi Ren, Jinzheng He, Xu Tan et al.

In pop music, accompaniments are usually played by multiple instruments (tracks) such as drum, bass, string and guitar, and can make a song more expressive and contagious by arranging together with its melody. Previous works usually generate multiple tracks separately and the music notes from different tracks not explicitly depend on each other, which hurts the harmony modeling. To improve harmony, in this paper, we propose a novel MUlti-track MIDI representation (MuMIDI), which enables simultaneous multi-track generation in a single sequence and explicitly models the dependency of the notes from different tracks. While this greatly improves harmony, unfortunately, it enlarges the sequence length and brings the new challenge of long-term music modeling. We further introduce two new techniques to address this challenge: 1) We model multiple note attributes (e.g., pitch, duration, velocity) of a musical note in one step instead of multiple steps, which can shorten the length of a MuMIDI sequence. 2) We introduce extra long-context as memory to capture long-term dependency in music. We call our system for pop music accompaniment generation as PopMAG. We evaluate PopMAG on multiple datasets (LMD, FreeMidi and CPMD, a private dataset of Chinese pop songs) with both subjective and objective metrics. The results demonstrate the effectiveness of PopMAG for multi-track harmony modeling and long-term context modeling. Specifically, PopMAG wins 42\%/38\%/40\% votes when comparing with ground truth musical pieces on LMD, FreeMidi and CPMD datasets respectively and largely outperforms other state-of-the-art music accompaniment generation models and multi-track MIDI representations in terms of subjective and objective metrics.

Jin Xu, Xu Tan, Yi Ren et al.

Speech synthesis (text to speech, TTS) and recognition (automatic speech recognition, ASR) are important speech tasks, and require a large amount of text and speech pairs for model training. However, there are more than 6,000 languages in the world and most languages are lack of speech training data, which poses significant challenges when building TTS and ASR systems for extremely low-resource languages. In this paper, we develop LRSpeech, a TTS and ASR system under the extremely low-resource setting, which can support rare languages with low data cost. LRSpeech consists of three key techniques: 1) pre-training on rich-resource languages and fine-tuning on low-resource languages; 2) dual transformation between TTS and ASR to iteratively boost the accuracy of each other; 3) knowledge distillation to customize the TTS model on a high-quality target-speaker voice and improve the ASR model on multiple voices. We conduct experiments on an experimental language (English) and a truly low-resource language (Lithuanian) to verify the effectiveness of LRSpeech. Experimental results show that LRSpeech 1) achieves high quality for TTS in terms of both intelligibility (more than 98% intelligibility rate) and naturalness (above 3.5 mean opinion score (MOS)) of the synthesized speech, which satisfy the requirements for industrial deployment, 2) achieves promising recognition accuracy for ASR, and 3) last but not least, uses extremely low-resource training data. We also conduct comprehensive analyses on LRSpeech with different amounts of data resources, and provide valuable insights and guidances for industrial deployment. We are currently deploying LRSpeech into a commercialized cloud speech service to support TTS on more rare languages.

Kaitao Song, Xu Tan, Jianfeng Lu

Neural machine translation (NMT) generates the next target token given as input the previous ground truth target tokens during training while the previous generated target tokens during inference, which causes discrepancy between training and inference as well as error propagation, and affects the translation accuracy. In this paper, we introduce an error correction mechanism into NMT, which corrects the error information in the previous generated tokens to better predict the next token. Specifically, we introduce two-stream self-attention from XLNet into NMT decoder, where the query stream is used to predict the next token, and meanwhile the content stream is used to correct the error information from the previous predicted tokens. We leverage scheduled sampling to simulate the prediction errors during training. Experiments on three IWSLT translation datasets and two WMT translation datasets demonstrate that our method achieves improvements over Transformer baseline and scheduled sampling. Further experimental analyses also verify the effectiveness of our proposed error correction mechanism to improve the translation quality.

Jinglin Liu, Yi Ren, Xu Tan et al.

Non-autoregressive translation (NAT) achieves faster inference speed but at the cost of worse accuracy compared with autoregressive translation (AT). Since AT and NAT can share model structure and AT is an easier task than NAT due to the explicit dependency on previous target-side tokens, a natural idea is to gradually shift the model training from the easier AT task to the harder NAT task. To smooth the shift from AT training to NAT training, in this paper, we introduce semi-autoregressive translation (SAT) as intermediate tasks. SAT contains a hyperparameter k, and each k value defines a SAT task with different degrees of parallelism. Specially, SAT covers AT and NAT as its special cases: it reduces to AT when k = 1 and to NAT when k = N (N is the length of target sentence). We design curriculum schedules to gradually shift k from 1 to N, with different pacing functions and number of tasks trained at the same time. We called our method as task-level curriculum learning for NAT (TCL-NAT). Experiments on IWSLT14 De-En, IWSLT16 En-De, WMT14 En-De and De-En datasets show that TCL-NAT achieves significant accuracy improvements over previous NAT baselines and reduces the performance gap between NAT and AT models to 1-2 BLEU points, demonstrating the effectiveness of our proposed method.

Yi Ren, Xu Tan, Tao Qin et al.

In this paper, we develop DeepSinger, a multi-lingual multi-singer singing voice synthesis (SVS) system, which is built from scratch using singing training data mined from music websites. The pipeline of DeepSinger consists of several steps, including data crawling, singing and accompaniment separation, lyrics-to-singing alignment, data filtration, and singing modeling. Specifically, we design a lyrics-to-singing alignment model to automatically extract the duration of each phoneme in lyrics starting from coarse-grained sentence level to fine-grained phoneme level, and further design a multi-lingual multi-singer singing model based on a feed-forward Transformer to directly generate linear-spectrograms from lyrics, and synthesize voices using Griffin-Lim. DeepSinger has several advantages over previous SVS systems: 1) to the best of our knowledge, it is the first SVS system that directly mines training data from music websites, 2) the lyrics-to-singing alignment model further avoids any human efforts for alignment labeling and greatly reduces labeling cost, 3) the singing model based on a feed-forward Transformer is simple and efficient, by removing the complicated acoustic feature modeling in parametric synthesis and leveraging a reference encoder to capture the timbre of a singer from noisy singing data, and 4) it can synthesize singing voices in multiple languages and multiple singers. We evaluate DeepSinger on our mined singing dataset that consists of about 92 hours data from 89 singers on three languages (Chinese, Cantonese and English). The results demonstrate that with the singing data purely mined from the Web, DeepSinger can synthesize high-quality singing voices in terms of both pitch accuracy and voice naturalness (footnote: Our audio samples are shown in https://speechresearch.github.io/deepsinger/.)

Chen Zhang, Xu Tan, Yi Ren et al.

Existing speech to speech translation systems heavily rely on the text of target language: they usually translate source language either to target text and then synthesize target speech from text, or directly to target speech with target text for auxiliary training. However, those methods cannot be applied to unwritten target languages, which have no written text or phoneme available. In this paper, we develop a translation system for unwritten languages, named as UWSpeech, which converts target unwritten speech into discrete tokens with a converter, and then translates source-language speech into target discrete tokens with a translator, and finally synthesizes target speech from target discrete tokens with an inverter. We propose a method called XL-VAE, which enhances vector quantized variational autoencoder (VQ-VAE) with cross-lingual (XL) speech recognition, to train the converter and inverter of UWSpeech jointly. Experiments on Fisher Spanish-English conversation translation dataset show that UWSpeech outperforms direct translation and VQ-VAE baseline by about 16 and 10 BLEU points respectively, which demonstrate the advantages and potentials of UWSpeech.

Kaitao Song, Hao Sun, Xu Tan et al.

While pre-training and fine-tuning, e.g., BERT~\citep{devlin2018bert}, GPT-2~\citep{radford2019language}, have achieved great success in language understanding and generation tasks, the pre-trained models are usually too big for online deployment in terms of both memory cost and inference speed, which hinders them from practical online usage. In this paper, we propose LightPAFF, a Lightweight Pre-training And Fine-tuning Framework that leverages two-stage knowledge distillation to transfer knowledge from a big teacher model to a lightweight student model in both pre-training and fine-tuning stages. In this way the lightweight model can achieve similar accuracy as the big teacher model, but with much fewer parameters and thus faster online inference speed. LightPAFF can support different pre-training methods (such as BERT, GPT-2 and MASS~\citep{song2019mass}) and be applied to many downstream tasks. Experiments on three language understanding tasks, three language modeling tasks and three sequence to sequence generation tasks demonstrate that while achieving similar accuracy with the big BERT, GPT-2 and MASS models, LightPAFF reduces the model size by nearly 5x and improves online inference speed by 5x-7x.

Yi Ren, Jinglin Liu, Xu Tan et al.

Non-autoregressive (NAR) models generate all the tokens of a sequence in parallel, resulting in faster generation speed compared to their autoregressive (AR) counterparts but at the cost of lower accuracy. Different techniques including knowledge distillation and source-target alignment have been proposed to bridge the gap between AR and NAR models in various tasks such as neural machine translation (NMT), automatic speech recognition (ASR), and text to speech (TTS). With the help of those techniques, NAR models can catch up with the accuracy of AR models in some tasks but not in some others. In this work, we conduct a study to understand the difficulty of NAR sequence generation and try to answer: (1) Why NAR models can catch up with AR models in some tasks but not all? (2) Why techniques like knowledge distillation and source-target alignment can help NAR models. Since the main difference between AR and NAR models is that NAR models do not use dependency among target tokens while AR models do, intuitively the difficulty of NAR sequence generation heavily depends on the strongness of dependency among target tokens. To quantify such dependency, we propose an analysis model called CoMMA to characterize the difficulty of different NAR sequence generation tasks. We have several interesting findings: 1) Among the NMT, ASR and TTS tasks, ASR has the most target-token dependency while TTS has the least. 2) Knowledge distillation reduces the target-token dependency in target sequence and thus improves the accuracy of NAR models. 3) Source-target alignment constraint encourages dependency of a target token on source tokens and thus eases the training of NAR models.

Xu Tan, Yichong Leng, Jiale Chen et al.

Multilingual neural machine translation (NMT) has recently been investigated from different aspects (e.g., pivot translation, zero-shot translation, fine-tuning, or training from scratch) and in different settings (e.g., rich resource and low resource, one-to-many, and many-to-one translation). This paper concentrates on a deep understanding of multilingual NMT and conducts a comprehensive study on a multilingual dataset with more than 20 languages. Our results show that (1) low-resource language pairs benefit much from multilingual training, while rich-resource language pairs may get hurt under limited model capacity and training with similar languages benefits more than dissimilar languages; (2) fine-tuning performs better than training from scratch in the one-to-many setting while training from scratch performs better in the many-to-one setting; (3) the bottom layers of the encoder and top layers of the decoder capture more language-specific information, and just fine-tuning these parts can achieve good accuracy for low-resource language pairs; (4) direct translation is better than pivot translation when the source language is similar to the target language (e.g., in the same language branch), even when the size of direct training data is much smaller; (5) given a fixed training data budget, it is better to introduce more languages into multilingual training for zero-shot translation.

Junliang Guo, Xu Tan, Linli Xu et al.

Non-autoregressive translation (NAT) models remove the dependence on previous target tokens and generate all target tokens in parallel, resulting in significant inference speedup but at the cost of inferior translation accuracy compared to autoregressive translation (AT) models. Considering that AT models have higher accuracy and are easier to train than NAT models, and both of them share the same model configurations, a natural idea to improve the accuracy of NAT models is to transfer a well-trained AT model to an NAT model through fine-tuning. However, since AT and NAT models differ greatly in training strategy, straightforward fine-tuning does not work well. In this work, we introduce curriculum learning into fine-tuning for NAT. Specifically, we design a curriculum in the fine-tuning process to progressively switch the training from autoregressive generation to non-autoregressive generation. Experiments on four benchmark translation datasets show that the proposed method achieves good improvement (more than $1$ BLEU score) over previous NAT baselines in terms of translation accuracy, and greatly speed up (more than $10$ times) the inference process over AT baselines.

Yingce Xia, Xu Tan, Fei Tian et al.

We Microsoft Research Asia made submissions to 11 language directions in the WMT19 news translation tasks. We won the first place for 8 of the 11 directions and the second place for the other three. Our basic systems are built on Transformer, back translation and knowledge distillation. We integrate several of our rececent techniques to enhance the baseline systems: multi-agent dual learning (MADL), masked sequence-to-sequence pre-training (MASS), neural architecture optimization (NAO), and soft contextual data augmentation (SCA).

Xu Tan, Yingce Xia, Lijun Wu et al.

The encoder-decoder based neural machine translation usually generates a target sequence token by token from left to right. Due to error propagation, the tokens in the right side of the generated sequence are usually of poorer quality than those in the left side. In this paper, we propose an efficient method to generate a sequence in both left-to-right and right-to-left manners using a single encoder and decoder, combining the advantages of both generation directions. Experiments on three translation tasks show that our method achieves significant improvements over conventional unidirectional approach. Compared with ensemble methods that train and combine two models with different generation directions, our method saves 50% model parameters and about 40% training time, and also improve inference speed.

Xu Tan, Jiale Chen, Di He et al.

Multilingual neural machine translation (NMT), which translates multiple languages using a single model, is of great practical importance due to its advantages in simplifying the training process, reducing online maintenance costs, and enhancing low-resource and zero-shot translation. Given there are thousands of languages in the world and some of them are very different, it is extremely burdensome to handle them all in a single model or use a separate model for each language pair. Therefore, given a fixed resource budget, e.g., the number of models, how to determine which languages should be supported by one model is critical to multilingual NMT, which, unfortunately, has been ignored by previous work. In this work, we develop a framework that clusters languages into different groups and trains one multilingual model for each cluster. We study two methods for language clustering: (1) using prior knowledge, where we cluster languages according to language family, and (2) using language embedding, in which we represent each language by an embedding vector and cluster them in the embedding space. In particular, we obtain the embedding vectors of all the languages by training a universal neural machine translation model. Our experiments on 23 languages show that the first clustering method is simple and easy to understand but leading to suboptimal translation accuracy, while the second method sufficiently captures the relationship among languages well and improves the translation accuracy for almost all the languages over baseline methods

Tianyu He, Xu Tan, Tao Qin

Neural machine translation (NMT) typically adopts the encoder-decoder framework. A good understanding of the characteristics and functionalities of the encoder and decoder can help to explain the pros and cons of the framework, and design better models for NMT. In this work, we conduct an empirical study on the encoder and the decoder in NMT, taking Transformer as an example. We find that 1) the decoder handles an easier task than the encoder in NMT, 2) the decoder is more sensitive to the input noise than the encoder, and 3) the preceding words/tokens in the decoder provide strong conditional information, which accounts for the two observations above. We hope those observations can shed light on the characteristics of the encoder and decoder and inspire future research on NMT.

Tianyu He, Jiale Chen, Xu Tan et al.

Neural machine translation on low-resource language is challenging due to the lack of bilingual sentence pairs. Previous works usually solve the low-resource translation problem with knowledge transfer in a multilingual setting. In this paper, we propose the concept of Language Graph and further design a novel graph distillation algorithm that boosts the accuracy of low-resource translations in the graph with forward and backward knowledge distillation. Preliminary experiments on the TED talks multilingual dataset demonstrate the effectiveness of our proposed method. Specifically, we improve the low-resource translation pair by more than 3.13 points in terms of BLEU score.

Jun Gao, Di He, Xu Tan et al.

We study an interesting problem in training neural network-based models for natural language generation tasks, which we call the \emph{representation degeneration problem}. We observe that when training a model for natural language generation tasks through likelihood maximization with the weight tying trick, especially with big training datasets, most of the learnt word embeddings tend to degenerate and be distributed into a narrow cone, which largely limits the representation power of word embeddings. We analyze the conditions and causes of this problem and propose a novel regularization method to address it. Experiments on language modeling and machine translation show that our method can largely mitigate the representation degeneration problem and achieve better performance than baseline algorithms.

Yichong Leng, Xu Tan, Tao Qin et al.

Unsupervised neural machine translation (NMT) has attracted a lot of attention recently. While state-of-the-art methods for unsupervised translation usually perform well between similar languages (e.g., English-German translation), they perform poorly between distant languages, because unsupervised alignment does not work well for distant languages. In this work, we introduce unsupervised pivot translation for distant languages, which translates a language to a distant language through multiple hops, and the unsupervised translation on each hop is relatively easier than the original direct translation. We propose a learning to route (LTR) method to choose the translation path between the source and target languages. LTR is trained on language pairs whose best translation path is available and is applied on the unseen language pairs for path selection. Experiments on 20 languages and 294 distant language pairs demonstrate the advantages of the unsupervised pivot translation for distant languages, as well as the effectiveness of the proposed LTR for path selection. Specifically, in the best case, LTR achieves an improvement of 5.58 BLEU points over the conventional direct unsupervised method.

Kaitao Song, Xu Tan, Tao Qin et al.

Pre-training and fine-tuning, e.g., BERT, have achieved great success in language understanding by transferring knowledge from rich-resource pre-training task to the low/zero-resource downstream tasks. Inspired by the success of BERT, we propose MAsked Sequence to Sequence pre-training (MASS) for the encoder-decoder based language generation tasks. MASS adopts the encoder-decoder framework to reconstruct a sentence fragment given the remaining part of the sentence: its encoder takes a sentence with randomly masked fragment (several consecutive tokens) as input, and its decoder tries to predict this masked fragment. In this way, MASS can jointly train the encoder and decoder to develop the capability of representation extraction and language modeling. By further fine-tuning on a variety of zero/low-resource language generation tasks, including neural machine translation, text summarization and conversational response generation (3 tasks and totally 8 datasets), MASS achieves significant improvements over the baselines without pre-training or with other pre-training methods. Specially, we achieve the state-of-the-art accuracy (37.5 in terms of BLEU score) on the unsupervised English-French translation, even beating the early attention-based supervised model.

Xu Tan, Yi Ren, Di He et al.

Multilingual machine translation, which translates multiple languages with a single model, has attracted much attention due to its efficiency of offline training and online serving. However, traditional multilingual translation usually yields inferior accuracy compared with the counterpart using individual models for each language pair, due to language diversity and model capacity limitations. In this paper, we propose a distillation-based approach to boost the accuracy of multilingual machine translation. Specifically, individual models are first trained and regarded as teachers, and then the multilingual model is trained to fit the training data and match the outputs of individual models simultaneously through knowledge distillation. Experiments on IWSLT, WMT and Ted talk translation datasets demonstrate the effectiveness of our method. Particularly, we show that one model is enough to handle multiple languages (up to 44 languages in our experiment), with comparable or even better accuracy than individual models.

Junliang Guo, Xu Tan, Di He et al.

Non-autoregressive translation (NAT) models, which remove the dependence on previous target tokens from the inputs of the decoder, achieve significantly inference speedup but at the cost of inferior accuracy compared to autoregressive translation (AT) models. Previous work shows that the quality of the inputs of the decoder is important and largely impacts the model accuracy. In this paper, we propose two methods to enhance the decoder inputs so as to improve NAT models. The first one directly leverages a phrase table generated by conventional SMT approaches to translate source tokens to target tokens, which are then fed into the decoder as inputs. The second one transforms source-side word embeddings to target-side word embeddings through sentence-level alignment and word-level adversary learning, and then feeds the transformed word embeddings into the decoder as inputs. Experimental results show our method largely outperforms the NAT baseline~\citep{gu2017non} by $5.11$ BLEU scores on WMT14 English-German task and $4.72$ BLEU scores on WMT16 English-Romanian task.

Chengyue Gong, Xu Tan, Di He et al.

Maximum-likelihood estimation (MLE) is widely used in sequence to sequence tasks for model training. It uniformly treats the generation/prediction of each target token as multi-class classification, and yields non-smooth prediction probabilities: in a target sequence, some tokens are predicted with small probabilities while other tokens are with large probabilities. According to our empirical study, we find that the non-smoothness of the probabilities results in low quality of generated sequences. In this paper, we propose a sentence-wise regularization method which aims to output smooth prediction probabilities for all the tokens in the target sequence. Our proposed method can automatically adjust the weights and gradients of each token in one sentence to ensure the predictions in a sequence uniformly well. Experiments on three neural machine translation tasks and one text summarization task show that our method outperforms conventional MLE loss on all these tasks and achieves promising BLEU scores on WMT14 English-German and WMT17 Chinese-English translation task.

Kaitao Song, Xu Tan, Furong Peng et al.

The encoder-decoder is the typical framework for Neural Machine Translation (NMT), and different structures have been developed for improving the translation performance. Transformer is one of the most promising structures, which can leverage the self-attention mechanism to capture the semantic dependency from global view. However, it cannot distinguish the relative position of different tokens very well, such as the tokens located at the left or right of the current token, and cannot focus on the local information around the current token either. To alleviate these problems, we propose a novel attention mechanism named Hybrid Self-Attention Network (HySAN) which accommodates some specific-designed masks for self-attention network to extract various semantic, such as the global/local information, the left/right part context. Finally, a squeeze gate is introduced to combine different kinds of SANs for fusion. Experimental results on three machine translation tasks show that our proposed framework outperforms the Transformer baseline significantly and achieves superior results over state-of-the-art NMT systems.

Chengyue Gong, Di He, Xu Tan et al.

Continuous word representation (aka word embedding) is a basic building block in many neural network-based models used in natural language processing tasks. Although it is widely accepted that words with similar semantics should be close to each other in the embedding space, we find that word embeddings learned in several tasks are biased towards word frequency: the embeddings of high-frequency and low-frequency words lie in different subregions of the embedding space, and the embedding of a rare word and a popular word can be far from each other even if they are semantically similar. This makes learned word embeddings ineffective, especially for rare words, and consequently limits the performance of these neural network models. In this paper, we develop a neat, simple yet effective way to learn \emph{FRequency-AGnostic word Embedding} (FRAGE) using adversarial training. We conducted comprehensive studies on ten datasets across four natural language processing tasks, including word similarity, language modeling, machine translation and text classification. Results show that with FRAGE, we achieve higher performance than the baselines in all tasks.

Lijun Wu, Xu Tan, Di He et al.

Neural machine translation usually adopts autoregressive models and suffers from exposure bias as well as the consequent error propagation problem. Many previous works have discussed the relationship between error propagation and the \emph{accuracy drop} (i.e., the left part of the translated sentence is often better than its right part in left-to-right decoding models) problem. In this paper, we conduct a series of analyses to deeply understand this problem and get several interesting findings. (1) The role of error propagation on accuracy drop is overstated in the literature, although it indeed contributes to the accuracy drop problem. (2) Characteristics of a language play a more important role in causing the accuracy drop: the left part of the translation result in a right-branching language (e.g., English) is more likely to be more accurate than its right part, while the right part is more accurate for a left-branching language (e.g., Japanese). Our discoveries are confirmed on different model structures including Transformer and RNN, and in other sequence generation tasks such as text summarization.

Kaitao Song, Xu Tan, Di He et al.

Encoder-decoder based Sequence to Sequence learning (S2S) has made remarkable progress in recent years. Different network architectures have been used in the encoder/decoder. Among them, Convolutional Neural Networks (CNN) and Self Attention Networks (SAN) are the prominent ones. The two architectures achieve similar performances but use very different ways to encode and decode context: CNN use convolutional layers to focus on the local connectivity of the sequence, while SAN uses self-attention layers to focus on global semantics. In this work we propose Double Path Networks for Sequence to Sequence learning (DPN-S2S), which leverage the advantages of both models by using double path information fusion. During the encoding step, we develop a double path architecture to maintain the information coming from different paths with convolutional layers and self-attention layers separately. To effectively use the encoded context, we develop a cross attention module with gating and use it to automatically pick up the information needed during the decoding step. By deeply integrating the two paths with cross attention, both types of information are combined and well exploited. Experiments show that our proposed method can significantly improve the performance of sequence to sequence learning over state-of-the-art systems.

Hany Hassan, Anthony Aue, Chang Chen et al.

Machine translation has made rapid advances in recent years. Millions of people are using it today in online translation systems and mobile applications in order to communicate across language barriers. The question naturally arises whether such systems can approach or achieve parity with human translations. In this paper, we first address the problem of how to define and accurately measure human parity in translation. We then describe Microsoft's machine translation system and measure the quality of its translations on the widely used WMT 2017 news translation task from Chinese to English. We find that our latest neural machine translation system has reached a new state-of-the-art, and that the translation quality is at human parity when compared to professional human translations. We also find that it significantly exceeds the quality of crowd-sourced non-professional translations.