Spoken language identification refers to the task of automatically predicting the spoken language in a given utterance. Conventionally, it is modeled as a speech-based language identification task. Prior techniques have been constrained to a single modality; however in the case of video data there is a wealth of other metadata that may be beneficial for this task. In this work, we propose MuSeLI, a Multimodal Spoken Language Identification method, which delves into the use of various metadata sources to enhance language identification. Our study reveals that metadata such as video title, description and geographic location provide substantial information to identify the spoken language of the multimedia recording. We conduct experiments using two diverse public datasets of YouTube videos, and obtain state-of-the-art results on the language identification task. We additionally conduct an ablation study that describes the distinct contribution of each modality for language recognition.

This paper proposes Virtuoso, a massively multilingual speech–text joint semi-supervised learning framework for text-to-speech synthesis (TTS) models. Existing multilingual TTS typically supports tens of languages, which are a small fraction of the thousands of languages in the world. One difficulty to scale multilingual TTS to hundreds of languages is collecting high-quality speech–text paired data in low-resource languages. This study extends Maestro, a speech–text joint pretraining framework for automatic speech recognition (ASR), to speech generation tasks. To train a TTS model from various types of speech and text data, different training schemes are designed to handle supervised (paired TTS and ASR data) and unsupervised (untranscribed speech and unspoken text) datasets. Experimental evaluation shows that 1) multilingual TTS models trained on Virtuoso can achieve significantly better naturalness and intelligibility than baseline ones in seen languages, and 2) they can synthesize reasonably intelligible and naturally sounding speech for unseen languages where no high-quality paired TTS data is available.

Much of text-to-speech research relies on human evaluation. This incurs heavy costs and slows down the development process, especially in heavily multilingual applications where recruiting and polling annotators can take weeks. We introduce SQuId (Speech Quality Identification), a multilingual naturalness prediction model trained on over a million ratings and tested in 65 locales—the largest effort of this type to date. The main insight is that training one model on many locales consistently surpasses mono-locale baselines. We show that the model outperforms a competitive baseline based on w2v-BERT and VoiceMOS by 50.0%. We then demonstrate the effectiveness of cross-locale transfer during fine-tuning and highlight its effect on zero-shot locales, for which there is no fine-tuning data. We highlight the role of non-linguistic effects such as sound artifacts in cross-locale transfer. Finally, we present the effect of model size and pre-training diversity with ablation experiments.

Recently, a number of approaches to train speech models by incorporating text into end-to-end models have been developed, with Maestro advancing state-of-the-art automatic speech recognition (ASR) and Speech Translation (ST) performance. In this paper, we expand our understanding of the resulting shared speech-text representations with two types of analyses. First we examine the limits of speech-free domain adaptation, finding that a corpus-specific duration model for speech-text alignment is the most important component for learning a shared speech-text representation. Second, we inspect the similarities between activations of unimodal (speech or text) encoders as compared to the activations of a shared encoder. We find that the shared encoder learns a more compact and overlapping speech-text representation than the uni-modal encoders. We hypothesize that this partially explains the effectiveness of the Maestro shared speech-text representations.

We present Mu$^2$SLAM, a multilingual sequence-to-sequence model pre-trained jointly on unlabeled speech, unlabeled text and supervised data spanning Automatic Speech Recognition (ASR), Automatic Speech Translation (AST) and Machine Translation (MT), in over 100 languages. By leveraging a quantized representation of speech as a target, Mu$^2$SLAM trains the speech-text models with a sequence-to-sequence masked denoising objective similar to T5 on the decoder and a masked language modeling objective (MLM) on the encoder, for both unlabeled speech and text, while utilizing the supervised tasks to improve cross-lingual and cross-modal representation alignment within the model. On CoVoST AST, Mu$^2$SLAM establishes a new state-of-the-art for models trained on public datasets, improving on xx-en translation over the previous best by 1.9 BLEU points and on en-xx translation by 1.1 BLEU points. On Voxpopuli ASR, our model matches the performance of an mSLAM model fine-tuned with an RNN-T decoder, despite using a relatively weaker Transformer decoder. On text understanding tasks, our model improves by more than 6% over mSLAM on XNLI, getting closer to the performance of mT5 models of comparable capacity on XNLI and TydiQA, paving the way towards a single model for all speech and text understanding tasks.

We present PronScribe, a novel method for phonemic transcription from speech and text input based on careful finetuning and adaptation of a massive, multilingual, multimodal speech-text pretrained model.We show that our model is capable of phonemically transcribing pronunciations of full utterances with accurate word boundaries in a variety of languages covering diverse phonological phenomena, achieving phoneme error rates in the vicinity of 1-2% which is comparable to human transcribers.We show that PronScribe can effectively learn this task from relatively little training data, making it attractive even in low-resource settings.It learns from text and speech simultaneously in a coherent way, and is better than previous models using speech, text or both.Additionally, the model's good transfer learning characteristics in multilingual settings can effectively boost performance for lower-resourced languages.

We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss the limitations revealed by self-auditing MADLAD-400, and the role data auditing had in the dataset creation process. We then train and release a 10.7B-parameter multilingual machine translation model on 250 billion tokens covering over 450 languages using publicly available data, and find that it is competitive with models that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot translation. We make the baseline models available to the research community.

Multilingual neural machine translation models are trained to maximize the likelihood of a mix of examples drawn from multiple language pairs.The dominant inductive bias applied to these models is a shared vocabulary and a shared set of parameters across languages; the inputs and labels corresponding to examples drawn from different language pairs might still reside in distinct subspaces.In this paper, we introduce multilingual crossover encoder-decoder (mXEncDec) to fuse language pairs at an instance level.Our approach interpolates instances from different language pairs into joint 'crossover examples' in order to encourage sharing input and output spaces across languages.To ensure better fusion of examples in multilingual settings, we propose several techniques to improve example interpolation across dissimilar languages under heavy data imbalance.Experiments on a large-scale WMT multilingual dataset demonstrate that our approach significantly improves quality on English-to-Many, Many-to-English and zero-shot translation tasks (from +0.5 BLEU up to +5.5 BLEU points).Results on code-switching sets demonstrate the capability of our approach to improve model generalization to out-of-distribution multilingual examples.We also conduct qualitative and quantitative representation comparisons to analyze the advantages of our approach at the representation level.

Biao Zhang, Ankur Bapna, Melvin Johnson, Ali Dabirmoghaddam, Naveen Arivazhagan, Orhan Firat. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.

Self-supervised training has shown promising gains in pretraining models and facilitating the downstream finetuning for speech recognition, like multilingual ASR. Most existing methods adopt a 2-stage scheme where the self-supervised loss is optimized in the first pretraining stage, and the standard supervised finetuning resumes in the second stage. In this paper, we propose an end-to-end (E2E) Joint Unsupervised and Supervised Training (JUST) method to combine the supervised RNN-T loss and the self-supervised contrastive and masked language modeling (MLM) losses. We validate its performance on the public dataset Multilingual LibriSpeech (MLS), which includes 8 languages and is extremely imbalanced. On MLS, we explore (1) JUST trained from scratch, and (2) JUST finetuned from a pretrained checkpoint. Experiments show that JUST can consistently outperform other existing state-of-the-art methods, and beat the monolingual baseline by a significant margin, demonstrating JUST’s capability of handling low-resource languages in multilingual ASR. Our average WER of all languages outperforms average monolingual baseline by 33.3%, and the state-of-the-art 2-stage XLSR by 32%. On low-resource languages like Polish, our WER is less than half of the monolingual baseline and even beats the supervised transfer learning method which uses external supervision.

Natural language understanding and generation models follow one of the two dominant architectural paradigms: language models (LMs) that process concatenated sequences in a single stack of layers, and encoder-decoder models (EncDec) that utilize separate layer stacks for input and output processing. In machine translation, EncDec has long been the favoured approach, but with few studies investigating the performance of LMs. In this work, we thoroughly examine the role of several architectural design choices on the performance of LMs on bilingual, (massively) multilingual and zero-shot translation tasks, under systematic variations of data conditions and model sizes. Our results show that: (i) Different LMs have different scaling properties, where architectural differences often have a significant impact on model performance at small scales, but the performance gap narrows as the number of parameters increases, (ii) Several design choices, including causal masking and language-modeling objectives for the source sequence, have detrimental effects on translation quality, and (iii) When paired with full-visible masking for source sequences, LMs could perform on par with EncDec on supervised bilingual and multilingual translation tasks, and improve greatly on zero-shot directions by facilitating the reduction of off-target translations.

We present Maestro, a self-supervised training method to unify representations learnt from speech and text modalities.Self-supervised learning from speech signals aims to learn the latent structure inherent in the signal, while self-supervised learning from text attempts to capture lexical information.Learning aligned representations from unpaired speech and text sequences is a challenging task.Previous work either implicitly enforced the representations learnt from these two modalities to be aligned in the latent space through multitasking and parameter sharing or explicitly through conversion of modalities via speech synthesis.While the former suffers from interference between the two modalities, the latter introduces additional complexity.In this paper, we propose Maestro, a novel algorithm to learn unified representations from both these modalities simultaneously that can transfer to diverse downstream tasks such as Automated Speech Recognition (ASR) and Speech Translation (ST).Maestro learns unified representations through sequence alignment, duration prediction and matching embeddings in the learned space through an aligned masked-language model loss.We establish a new state-of-the-art (SOTA) on VoxPopuli multilingual ASR with a 8% relative reduction in Word Error Rate (WER), multidomain SpeechStew ASR (3.7% relative) and 21 languages to English multilingual ST on CoVoST 2 with an improvement of 2.8 BLEU averaged over 21 languages.

We introduce XTREME-S, a new benchmark to evaluate universal cross-lingual speech representations in many languages.XTREME-S covers four task families: speech recognition, classification, speech-to-text translation and retrieval.Covering 102 languages from 10+ language families, 3 different domains and 4 task families, XTREME-S aims to simplify multilingual speech representation evaluation, as well as catalyze research in "universal" speech representation learning.This paper describes the new benchmark and establishes the first speech-only and speechtext baselines using XLS-R and mSLAM on all downstream tasks.We motivate the design choices and detail how to use the benchmark.Datasets and fine-tuning scripts are made easily accessible through the HuggingFace platform. 1

End-to-end speech-to-speech translation (S2ST) without relying on intermediate text representations is a rapidly emerging frontier of research.Recent works have demonstrated that the performance of such direct S2ST systems is approaching that of conventional cascade S2ST when trained on comparable datasets.However, in practice, the performance of direct S2ST is bounded by the availability of paired S2ST training data.In this work, we explore multiple approaches for leveraging much more widely available unsupervised and weakly-supervised speech and text data to improve the performance of direct S2ST based on Translatotron 2. With our most effective approaches, the average translation quality of direct S2ST on 21 language pairs on the CVSS-C corpus is improved by +13.6 BLEU (or +113% relatively), as compared to the previous state-of-the-art trained without additional data.The improvements on low-resource language are even more significant (+398% relatively on average).Our comparative studies suggest future research directions for S2ST and speech representation learning.

Training state-of-the-art Automated Speech Recognition (ASR) models typically requires a substantial amount of transcribed speech. In this work, we demonstrate that a modality-matched joint speech and text model introduced in [1] can be leveraged to train a massively multilingual ASR model without any supervised (manually transcribed) speech for some languages. This paper explores the use of jointly learnt speech and text representations in a massively multilingual, zero supervised speech, real-world setting to expand the set of languages covered by ASR with only unlabeled speech and text in the target languages. Using the FLEURS dataset, we define the task to cover 102 languages, where transcribed speech is available in 52 of these languages and can be used to improve end-to-end ASR quality on the remaining 50. First, we show that by combining speech representations with byte-level text representations and use of language embeddings, we can dramatically reduce the Character Error Rate (CER) on languages with no supervised speech from 64.8% to 30.8%, a relative reduction of 53%. Second, using a subset of South Asian languages we show that Maestro-U can promote knowledge transfer from languages with supervised speech even when there is limited to no graphemic overlap. Overall, Maestro-U closes the gap to oracle performance by 68.5% relative and reduces the CER of 19 languages below 15%.

We introduce FLEURS, the Few-shot Learning Evaluation of Universal Representations of Speech benchmark. FLEURS is an n-way parallel speech dataset in 102 languages built on top of the machine translation FLoRes-101 benchmark, with approximately 12 hours of speech supervision per language. FLEURS can be used for a variety of speech tasks, including Automatic Speech Recognition (ASR), Speech Language Identification (Speech LangID), Speech-Text Retrieval. In this paper, we provide baselines for the tasks based on multilingual pre-trained models like speech-only w2v-BERT [1] and speech-text multimodal mSLAM [2]. The goal of FLEURS is to enable speech technology in more languages and catalyze research in low-resource speech understanding.1.

We present JOIST, an algorithm to train a streaming, cascaded, encoder end-to-end (E2E) model with both speech-text paired inputs, and text-only unpaired inputs. Unlike previous works, we explore joint training with both modalities, rather than pre-training and fine-tuning. In addition, we explore JOIST using a streaming E2E model with an order of magnitude more data, which are also novelties compared to previous works. Through a series of ablation studies, we explore different types of text modeling, including how to model the length of the text sequence and the appropriate text subword unit representation. We find that best text representation for JOIST improves WER across a variety of search and rare-word test sets by 4-14% relative, compared to a model not trained with text. In addition, we quantitatively show that JOIST maintains streaming capabilities, which is important for good user-level experience.

Abstract With the success of large-scale pre-training and multilingual modeling in Natural Language Processing (NLP), recent years have seen a proliferation of large, Web-mined text datasets covering hundreds of languages. We manually audit the quality of 205 language-specific corpora released with five major public datasets (CCAligned, ParaCrawl, WikiMatrix, OSCAR, mC4). Lower-resource corpora have systematic issues: At least 15 corpora have no usable text, and a significant fraction contains less than 50% sentences of acceptable quality. In addition, many are mislabeled or use nonstandard/ambiguous language codes. We demonstrate that these issues are easy to detect even for non-proficient speakers, and supplement the human audit with automatic analyses. Finally, we recommend techniques to evaluate and improve multilingual corpora and discuss potential risks that come with low-quality data releases.

The recently proposed massively multilingual neural machine translation (NMT) system has been shown to be capable of translating over 100 languages to and from English within a single model (Aharoni, Johnson, and Firat 2019). Its improved translation performance on low resource languages hints at potential cross-lingual transfer capability for downstream tasks. In this paper, we evaluate the cross-lingual effectiveness of representations from the encoder of a massively multilingual NMT model on 5 downstream classification and sequence labeling tasks covering a diverse set of over 50 languages. We compare against a strong baseline, multilingual BERT (mBERT) (Devlin et al. 2018), in different cross-lingual transfer learning scenarios and show gains in zero-shot transfer in 4 out of these 5 tasks.

Aditya Siddhant, Ankur Bapna, Yuan Cao, Orhan Firat, Mia Chen, Sneha Kudugunta, Naveen Arivazhagan, Yonghui Wu. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020.

Large text corpora are increasingly important for a wide variety of Natural Language Processing (NLP) tasks, and automatic language identification (LangID) is a core technology needed to collect such datasets in a multilingual context.LangID is largely treated as solved in the literature, with models reported that achieve over 90% average F1 on as many as 1,366 languages.We train LangID models on up to 1,629 languages with comparable quality on held-out test sets, but find that human-judged LangID accuracy for web-crawl text corpora created using these models is only around 5% for many lower-resource languages, suggesting a need for more robust evaluation.Further analysis revealed a variety of error modes, arising from domain mismatch, class imbalance, language similarity, and insufficiently expressive models.We propose two classes of techniques to mitigate these errors: wordlist-based tunable-precision filters (for which we release curated lists in about 500 languages) and transformer-based semi-supervised LangID models, which increase median dataset precision from 5.5% to 71.2%.These techniques enable us to create an initial data set covering 100K or more relatively clean sentences in each of 500+ languages, paving the way towards a 1,000-language web text corpus.

Ankur Bapna, Orhan Firat. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Sneha Kudugunta, Ankur Bapna, Isaac Caswell, Orhan Firat. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Multilingual end-to-end (E2E) models have shown great promise in expansion of automatic speech recognition (ASR) coverage of the world's languages.They have shown improvement over monolingual systems, and have simplified training and serving by eliminating language-specific acoustic, pronunciation, and language models.This work presents an E2E multilingual system which is equipped to operate in low-latency interactive applications, as well as handle a key challenge of real world data: the imbalance in training data across languages.Using nine Indic languages, we compare a variety of techniques, and find that a combination of conditioning on a language vector and training language-specific adapter layers produces the best model.The resulting E2E multilingual model achieves a lower word error rate (WER) than both monolingual E2E models (eight of nine languages) and monolingual conventional systems (all nine languages).

Scaling up deep neural network capacity has been known as an effective approach to improving model quality for several different machine learning tasks. In many cases, increasing model capacity beyond the memory limit of a single accelerator has required developing special algorithms or infrastructure. These solutions are often architecture-specific and do not transfer to other machine learning tasks. To address the need for efficient and task-independent model parallelism, we introduce TensorPipe, a pipeline parallelism library that allows scaling any network that can be expressed as a sequence of layers. By pipelining different sub-sequences of layers on separate accelerators, TensorPipe provides the flexibility of scaling a variety of different networks to gigantic sizes efficiently. Moreover, TensorPipe utilizes a novel batch-splitting pipelining algorithm, resulting in almost linear speedup when a model is partitioned across multiple accelerators. We demonstrate the advantages of TensorPipe by training large-scale neural networks on two different tasks with distinct network architectures: (i)Image Classification: We train a 557-million-parameter AmoebaNet model and attain a top-1 accuracy of 84.4% on ImageNet-2012, (ii)Multilingual Neural Machine Translation: We train a single 6-billion-parameter, 128-layer Transformer model on a corpus spanning over 100 languages and achieve better quality than all bilingual models.

Mia Xu Chen, Orhan Firat, Ankur Bapna, Melvin Johnson, Wolfgang Macherey, George Foster, Llion Jones, Mike Schuster, Noam Shazeer, Niki Parmar, Ashish Vaswani, Jakob Uszkoreit, Lukasz Kaiser, Zhifeng Chen, Yonghui Wu, Macduff Hughes. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2018.

While current state-of-the-art NMT models, such as RNN seq2seq and Transformers, possess a large number of parameters, they are still shallow in comparison to convolutional models used for both text and vision applications.In this work we attempt to train significantly (2-3x) deeper Transformer and Bi-RNN encoders for machine translation.We propose a simple modification to the attention mechanism that eases the optimization of deeper models, and results in consistent gains of 0.7-1.1 BLEU on the benchmark WMT'14 English-German and WMT'15 Czech-English tasks for both architectures.

Translating characters instead of words or word-fragments has the potential to simplify the processing pipeline for neural machine translation (NMT), and improve results by eliminating hyper-parameters and manual feature engineering.However, it results in longer sequences in which each symbol contains less information, creating both modeling and computational challenges.In this paper, we show that the modeling problem can be solved by standard sequence-to-sequence architectures of sufficient depth, and that deep models operating at the character level outperform identical models operating over word fragments.This result implies that alternative architectures for handling character input are better viewed as methods for reducing computation time than as improved ways of modeling longer sequences.From this perspective, we evaluate several techniques for characterlevel NMT, verify that they do not match the performance of our deep character baseline model, and evaluate the performance versus computation time tradeoffs they offer.Within this framework, we also perform the first evaluation for NMT of conditional computation over time, in which the model learns which timesteps can be skipped, rather than having them be dictated by a fixed schedule specified before training begins.

State-of-the-art slot filling models for goal-oriented human/machine conversational language understanding systems rely on deep learning methods. While multi-task training of such models alleviates the need for large in-domain annotated datasets, bootstrapping a semantic parsing model for a new domain using only the semantic frame, such as the back-end API or knowledge graph schema, is still one of the holy grail tasks of language understanding for dialogue systems. This paper proposes a deep learning based approach that can utilize only the slot description in context without the need for any labeled or unlabeled in-domain examples, to quickly bootstrap a new domain. The main idea of this paper is to leverage the encoding of the slot names and descriptions within a multi-task deep learned slot filling model, to implicitly align slots across domains. The proposed approach is promising for solving the domain scaling problem and eliminating the need for any manually annotated data or explicit schema alignment. Furthermore, our experiments on multiple domains show that this approach results in significantly better slot-filling performance when compared to using only in-domain data, especially in the low data regime.

Spoken Language Understanding (SLU) is a key component of goal oriented dialogue systems that would parse user utterances into semantic frame representations.Traditionally SLU does not utilize the dialogue history beyond the previous system turn and contextual ambiguities are resolved by the downstream components.In this paper, we explore novel approaches for modeling dialogue context in a recurrent neural network (RNN) based language understanding system.We propose the Sequential Dialogue Encoder Network, that allows encoding context from the dialogue history in chronological order.We compare the performance of our proposed architecture with two context models, one that uses just the previous turn context and another that encodes dialogue context in a memory network, but loses the order of utterances in the dialogue history.Experiments with a multi-domain dialogue dataset demonstrate that the proposed architecture results in reduced semantic frame error rates.