Mingxuan Wang

Rong Ye, Mingxuan Wang, Lei Li · bytedance, cmu

How can we learn unified representations for spoken utterances and their written text? Learning similar representations for semantically similar speech and text is important for speech translation. To this end, we propose ConST, a cross-modal contrastive learning method for end-to-end speech-to-text translation. We evaluate ConST and a variety of previous baselines on a popular benchmark MuST-C. Experiments show that the proposed ConST consistently outperforms the previous methods on, and achieves an average BLEU of 29.4. The analysis further verifies that ConST indeed closes the representation gap of different modalities -- its learned representation improves the accuracy of cross-modal speech-text retrieval from 4% to 88%. Code and models are available at https://github.com/ReneeYe/ConST.

Qingkai Fang, Rong Ye, Lei Li et al. · bytedance, cmu

How to learn a better speech representation for end-to-end speech-to-text translation (ST) with limited labeled data? Existing techniques often attempt to transfer powerful machine translation (MT) capabilities to ST, but neglect the representation discrepancy across modalities. In this paper, we propose the Speech-TExt Manifold Mixup (STEMM) method to calibrate such discrepancy. Specifically, we mix up the representation sequences of different modalities, and take both unimodal speech sequences and multimodal mixed sequences as input to the translation model in parallel, and regularize their output predictions with a self-learning framework. Experiments on MuST-C speech translation benchmark and further analysis show that our method effectively alleviates the cross-modal representation discrepancy, and achieves significant improvements over a strong baseline on eight translation directions.

Xian Qian, Kai Hu, Jiaqiang Wang et al. · bytedance

This report describes our VolcTrans system for the WMT22 shared task on large-scale multilingual machine translation. We participated in the unconstrained track which allows the use of external resources. Our system is a transformerbased multilingual model trained on data from multiple sources including the public training set from the data track, NLLB data provided by Meta AI, self-collected parallel corpora, and pseudo bitext from back-translation. A series of heuristic rules clean both bilingual and monolingual texts. On the official test set, our system achieves 17.3 BLEU, 21.9 spBLEU, and 41.9 chrF2++ on average over all language pairs. The average inference speed is 11.5 sentences per second using a single Nvidia Tesla V100 GPU. Our code and trained models are available at https://github.com/xian8/wmt22

Qianqian Dong, Zhiying Huang, Qiao Tian et al.

We propose PolyVoice, a language model-based framework for speech-to-speech translation (S2ST) system. Our framework consists of two language models: a translation language model and a speech synthesis language model. We use discretized speech units, which are generated in a fully unsupervised way, and thus our framework can be used for unwritten languages. For the speech synthesis part, we adopt the existing VALL-E X approach and build a unit-based audio language model. This grants our framework the ability to preserve the voice characteristics and the speaking style of the original speech. We examine our system on Chinese $\rightarrow$ English and English $\rightarrow$ Spanish pairs. Experimental results show that our system can generate speech with high translation quality and audio quality. Speech samples are available at https://speechtranslation.github.io/polyvoice.

Xuxin Cheng, Qianqian Dong, Fengpeng Yue et al.

How to solve the data scarcity problem for end-to-end speech-to-text translation (ST)? It's well known that data augmentation is an efficient method to improve performance for many tasks by enlarging the dataset. In this paper, we propose Mix at three levels for Speech Translation (M^3ST) method to increase the diversity of the augmented training corpus. Specifically, we conduct two phases of fine-tuning based on a pre-trained model using external machine translation (MT) data. In the first stage of fine-tuning, we mix the training corpus at three levels, including word level, sentence level and frame level, and fine-tune the entire model with mixed data. At the second stage of fine-tuning, we take both original speech sequences and original text sequences in parallel into the model to fine-tune the network, and use Jensen-Shannon divergence to regularize their outputs. Experiments on MuST-C speech translation benchmark and analysis show that M^3ST outperforms current strong baselines and achieves state-of-the-art results on eight directions with an average BLEU of 29.9.

Kexin Wang, Yunlong Zhao, Qianqian Dong et al.

As a subjective metric to evaluate the quality of synthesized speech, Mean opinion score~(MOS) usually requires multiple annotators to score the same speech. Such an annotation approach requires a lot of manpower and is also time-consuming. MOS prediction model for automatic evaluation can significantly reduce labor cost. In previous works, it is difficult to accurately rank the quality of speech when the MOS scores are close. However, in practical applications, it is more important to correctly rank the quality of synthesis systems or sentences than simply predicting MOS scores. Meanwhile, as each annotator scores multiple audios during annotation, the score is probably a relative value based on the first or the first few speech scores given by the annotator. Motivated by the above two points, we propose a general framework for MOS prediction based on pair comparison (MOSPC), and we utilize C-Mixup algorithm to enhance the generalization performance of MOSPC. The experiments on BVCC and VCC2018 show that our framework outperforms the baselines on most of the correlation coefficient metrics, especially on the metric KTAU related to quality ranking. And our framework also surpasses the strong baseline in ranking accuracy on each fine-grained segment. These results indicate that our framework contributes to improving the ranking accuracy of speech quality.

Dong Zhang, Rong Ye, Tom Ko et al.

How can speech-to-text translation (ST) perform as well as machine translation (MT)? The key point is to bridge the modality gap between speech and text so that useful MT techniques can be applied to ST. Recently, the approach of representing speech with unsupervised discrete units yields a new way to ease the modality problem. This motivates us to propose Discrete Unit Back-translation (DUB) to answer two questions: (1) Is it better to represent speech with discrete units than with continuous features in direct ST? (2) How much benefit can useful MT techniques bring to ST? With DUB, the back-translation technique can successfully be applied on direct ST and obtains an average boost of 5.5 BLEU on MuST-C En-De/Fr/Es. In the low-resource language scenario, our method achieves comparable performance to existing methods that rely on large-scale external data. Code and models are available at https://github.com/0nutation/DUB.

Yaoming Zhu, Liwei Wu, Shanbo Cheng et al.

The punctuation restoration task aims to correctly punctuate the output transcriptions of automatic speech recognition systems. Previous punctuation models, either using text only or demanding the corresponding audio, tend to be constrained by real scenes, where unpunctuated sentences are a mixture of those with and without audio. This paper proposes a unified multimodal punctuation restoration framework, named UniPunc, to punctuate the mixed sentences with a single model. UniPunc jointly represents audio and non-audio samples in a shared latent space, based on which the model learns a hybrid representation and punctuates both kinds of samples. We validate the effectiveness of the UniPunc on real-world datasets, which outperforms various strong baselines (e.g. BERT, MuSe) by at least 0.8 overall F1 scores, making a new state-of-the-art. Extensive experiments show that UniPunc's design is a pervasive solution: by grafting onto previous models, UniPunc enables them to punctuate on the mixed corpus. Our code is available at github.com/Yaoming95/UniPunc

Qianqian Dong, Yaoming Zhu, Mingxuan Wang et al.

How to find proper moments to generate partial sentence translation given a streaming speech input? Existing approaches waiting-and-translating for a fixed duration often break the acoustic units in speech, since the boundaries between acoustic units in speech are not even. In this paper, we propose MoSST, a simple yet effective method for translating streaming speech content. Given a usually long speech sequence, we develop an efficient monotonic segmentation module inside an encoder-decoder model to accumulate acoustic information incrementally and detect proper speech unit boundaries for the input in speech translation task. Experiments on multiple translation directions of the MuST-C dataset show that MoSST outperforms existing methods and achieves the best trade-off between translation quality (BLEU) and latency. Our code is available at https://github.com/dqqcasia/mosst.

Jianze Liang, Chengqi Zhao, Mingxuan Wang et al.

Neural machine translation often adopts the fine-tuning approach to adapt to specific domains. However, nonrestricted fine-tuning can easily degrade on the general domain and over-fit to the target domain. To mitigate the issue, we propose Prune-Tune, a novel domain adaptation method via gradual pruning. It learns tiny domain-specific sub-networks during fine-tuning on new domains. Prune-Tune alleviates the over-fitting and the degradation problem without model modification. Furthermore, Prune-Tune is able to sequentially learn a single network with multiple disjoint domain-specific sub-networks for multiple domains. Empirical experiment results show that Prune-Tune outperforms several strong competitors in the target domain test set without sacrificing the quality on the general domain in both single and multi-domain settings. The source code and data are available at https://github.com/ohlionel/Prune-Tune.

Qianqian Dong, Mingxuan Wang, Hao Zhou et al.

Speech-to-text translation (ST), which directly translates the source language speech to the target language text, has attracted intensive attention recently. However, the combination of speech recognition and machine translation in a single model poses a heavy burden on the direct cross-modal cross-lingual mapping. To reduce the learning difficulty, we propose COnSecutive Transcription and Translation (COSTT), an integral approach for speech-to-text translation. The key idea is to generate source transcript and target translation text with a single decoder. It benefits the model training so that additional large parallel text corpus can be fully exploited to enhance the speech translation training. Our method is verified on three mainstream datasets, including Augmented LibriSpeech English-French dataset, IWSLT2018 English-German dataset, and TED English-Chinese dataset. Experiments show that our proposed COSTT outperforms or on par with the previous state-of-the-art methods on the three datasets. We have released our code at \url{https://github.com/dqqcasia/st}.

Qianqian Dong, Rong Ye, Mingxuan Wang et al.

An end-to-end speech-to-text translation (ST) takes audio in a source language and outputs the text in a target language. Existing methods are limited by the amount of parallel corpus. Can we build a system to fully utilize signals in a parallel ST corpus? We are inspired by human understanding system which is composed of auditory perception and cognitive processing. In this paper, we propose Listen-Understand-Translate, (LUT), a unified framework with triple supervision signals to decouple the end-to-end speech-to-text translation task. LUT is able to guide the acoustic encoder to extract as much information from the auditory input. In addition, LUT utilizes a pre-trained BERT model to enforce the upper encoder to produce as much semantic information as possible, without extra data. We perform experiments on a diverse set of speech translation benchmarks, including Librispeech English-French, IWSLT English-German and TED English-Chinese. Our results demonstrate LUT achieves the state-of-the-art performance, outperforming previous methods. The code is available at https://github.com/dqqcasia/st.

Tiantian Fan, Lingjun Liu, Yu Yue et al.

Recently, test-time scaling Large Language Models (LLMs) have demonstrated exceptional reasoning capabilities across scientific and professional tasks by generating long chains-of-thought (CoT). As a crucial component for developing these reasoning models, reinforcement learning (RL), exemplified by Proximal Policy Optimization (PPO) and its variants, allows models to learn through trial and error. However, PPO can be time-consuming due to its inherent on-policy nature, which is further exacerbated by increasing response lengths. In this work, we propose Truncated Proximal Policy Optimization (T-PPO), a novel extension to PPO that improves training efficiency by streamlining policy update and length-restricted response generation. T-PPO mitigates the issue of low hardware utilization, an inherent drawback of fully synchronized long-generation procedures, where resources often sit idle during the waiting periods for complete rollouts. Our contributions are two-folds. First, we propose Extended Generalized Advantage Estimation (EGAE) for advantage estimation derived from incomplete responses while maintaining the integrity of policy learning. Second, we devise a computationally optimized mechanism that allows for the independent optimization of the policy and value models. By selectively filtering prompt and truncated tokens, this mechanism reduces redundant computations and accelerates the training process without sacrificing convergence performance. We demonstrate the effectiveness and efficacy of T-PPO on AIME 2024 with a 32B base model. The experimental results show that T-PPO improves the training efficiency of reasoning LLMs by up to 2.5x and outperforms its existing competitors.

Mingxuan Wang, Zhengdong Lu, Jie Zhou et al.

Deep Neural Networks (DNNs) have provably enhanced the state-of-the-art Neural Machine Translation (NMT) with their capability in modeling complex functions and capturing complex linguistic structures. However NMT systems with deep architecture in their encoder or decoder RNNs often suffer from severe gradient diffusion due to the non-linear recurrent activations, which often make the optimization much more difficult. To address this problem we propose novel linear associative units (LAU) to reduce the gradient propagation length inside the recurrent unit. Different from conventional approaches (LSTM unit and GRU), LAUs utilizes linear associative connections between input and output of the recurrent unit, which allows unimpeded information flow through both space and time direction. The model is quite simple, but it is surprisingly effective. Our empirical study on Chinese-English translation shows that our model with proper configuration can improve by 11.7 BLEU upon Groundhog and the best reported results in the same setting. On WMT14 English-German task and a larger WMT14 English-French task, our model achieves comparable results with the state-of-the-art.

Mingxuan Wang, Zhengdong Lu, Hang Li et al.

We propose to enhance the RNN decoder in a neural machine translator (NMT) with external memory, as a natural but powerful extension to the state in the decoding RNN. This memory-enhanced RNN decoder is called \textsc{MemDec}. At each time during decoding, \textsc{MemDec} will read from this memory and write to this memory once, both with content-based addressing. Unlike the unbounded memory in previous work\cite{RNNsearch} to store the representation of source sentence, the memory in \textsc{MemDec} is a matrix with pre-determined size designed to better capture the information important for the decoding process at each time step. Our empirical study on Chinese-English translation shows that it can improve by $4.8$ BLEU upon Groundhog and $5.3$ BLEU upon on Moses, yielding the best performance achieved with the same training set.

Mingxuan Wang, Zhengdong Lu, Hang Li et al.

We propose a novel convolutional architecture, named $gen$CNN, for word sequence prediction. Different from previous work on neural network-based language modeling and generation (e.g., RNN or LSTM), we choose not to greedily summarize the history of words as a fixed length vector. Instead, we use a convolutional neural network to predict the next word with the history of words of variable length. Also different from the existing feedforward networks for language modeling, our model can effectively fuse the local correlation and global correlation in the word sequence, with a convolution-gating strategy specifically designed for the task. We argue that our model can give adequate representation of the history, and therefore can naturally exploit both the short and long range dependencies. Our model is fast, easy to train, and readily parallelized. Our extensive experiments on text generation and $n$-best re-ranking in machine translation show that $gen$CNN outperforms the state-of-the-arts with big margins.

Mingxuan Wang, Zhengdong Lu, Hang Li et al.

Many tasks in natural language processing, ranging from machine translation to question answering, can be reduced to the problem of matching two sentences or more generally two short texts. We propose a new approach to the problem, called Deep Match Tree (DeepMatch$_{tree}$), under a general setting. The approach consists of two components, 1) a mining algorithm to discover patterns for matching two short-texts, defined in the product space of dependency trees, and 2) a deep neural network for matching short texts using the mined patterns, as well as a learning algorithm to build the network having a sparse structure. We test our algorithm on the problem of matching a tweet and a response in social media, a hard matching problem proposed in [Wang et al., 2013], and show that DeepMatch$_{tree}$ can outperform a number of competitor models including one without using dependency trees and one based on word-embedding, all with large margins

Fandong Meng, Zhengdong Lu, Mingxuan Wang et al.

The recently proposed neural network joint model (NNJM) (Devlin et al., 2014) augments the n-gram target language model with a heuristically chosen source context window, achieving state-of-the-art performance in SMT. In this paper, we give a more systematic treatment by summarizing the relevant source information through a convolutional architecture guided by the target information. With different guiding signals during decoding, our specifically designed convolution+gating architectures can pinpoint the parts of a source sentence that are relevant to predicting a target word, and fuse them with the context of entire source sentence to form a unified representation. This representation, together with target language words, are fed to a deep neural network (DNN) to form a stronger NNJM. Experiments on two NIST Chinese-English translation tasks show that the proposed model can achieve significant improvements over the previous NNJM by up to +1.08 BLEU points on average