Antoine Laurent

Sameer Khurana, Antoine Laurent, James Glass · mit

We propose the SAMU-XLSR: Semantically-Aligned Multimodal Utterance-level Cross-Lingual Speech Representation learning framework. Unlike previous works on speech representation learning, which learns multilingual contextual speech embedding at the resolution of an acoustic frame (10-20ms), this work focuses on learning multimodal (speech-text) multilingual speech embedding at the resolution of a sentence (5-10s) such that the embedding vector space is semantically aligned across different languages. We combine state-of-the-art multilingual acoustic frame-level speech representation learning model XLS-R with the Language Agnostic BERT Sentence Embedding (LaBSE) model to create an utterance-level multimodal multilingual speech encoder SAMU-XLSR. Although we train SAMU-XLSR with only multilingual transcribed speech data, cross-lingual speech-text and speech-speech associations emerge in its learned representation space. To substantiate our claims, we use SAMU-XLSR speech encoder in combination with a pre-trained LaBSE text sentence encoder for cross-lingual speech-to-text translation retrieval, and SAMU-XLSR alone for cross-lingual speech-to-speech translation retrieval. We highlight these applications by performing several cross-lingual text and speech translation retrieval tasks across several datasets.

Nauman Dawalatabad, Sameer Khurana, Antoine Laurent et al. · mit

Pseudo-label (PL) filtering forms a crucial part of Self-Training (ST) methods for unsupervised domain adaptation. Dropout-based Uncertainty-driven Self-Training (DUST) proceeds by first training a teacher model on source domain labeled data. Then, the teacher model is used to provide PLs for the unlabeled target domain data. Finally, we train a student on augmented labeled and pseudo-labeled data. The process is iterative, where the student becomes the teacher for the next DUST iteration. A crucial step that precedes the student model training in each DUST iteration is filtering out noisy PLs that could lead the student model astray. In DUST, we proposed a simple, effective, and theoretically sound PL filtering strategy based on the teacher model's uncertainty about its predictions on unlabeled speech utterances. We estimate the model's uncertainty by computing disagreement amongst multiple samples drawn from the teacher model during inference by injecting noise via dropout. In this work, we show that DUST's PL filtering, as initially used, may fail under severe source and target domain mismatch. We suggest several approaches to eliminate or alleviate this issue. Further, we bring insights from the research in neural network model calibration to DUST and show that a well-calibrated model correlates strongly with a positive outcome of the DUST PL filtering step.

Sameer Khurana, Nauman Dawalatabad, Antoine Laurent et al. · mit

Research in multilingual speech-to-text translation is topical. Having a single model that supports multiple translation tasks is desirable. The goal of this work it to improve cross-lingual transfer learning in multilingual speech-to-text translation via semantic knowledge distillation. We show that by initializing the encoder of the encoder-decoder sequence-to-sequence translation model with SAMU-XLS-R, a multilingual speech transformer encoder trained using multi-modal (speech-text) semantic knowledge distillation, we achieve significantly better cross-lingual task knowledge transfer than the baseline XLS-R, a multilingual speech transformer encoder trained via self-supervised learning. We demonstrate the effectiveness of our approach on two popular datasets, namely, CoVoST-2 and Europarl. On the 21 translation tasks of the CoVoST-2 benchmark, we achieve an average improvement of 12.8 BLEU points over the baselines. In the zero-shot translation scenario, we achieve an average gain of 18.8 and 11.9 average BLEU points on unseen medium and low-resource languages. We make similar observations on Europarl speech translation benchmark.

Marcely Zanon Boito, John Ortega, Hugo Riguidel et al. · meta-ai

This paper describes the ON-TRAC Consortium translation systems developed for two challenge tracks featured in the Evaluation Campaign of IWSLT 2022: low-resource and dialect speech translation. For the Tunisian Arabic-English dataset (low-resource and dialect tracks), we build an end-to-end model as our joint primary submission, and compare it against cascaded models that leverage a large fine-tuned wav2vec 2.0 model for ASR. Our results show that in our settings pipeline approaches are still very competitive, and that with the use of transfer learning, they can outperform end-to-end models for speech translation (ST). For the Tamasheq-French dataset (low-resource track) our primary submission leverages intermediate representations from a wav2vec 2.0 model trained on 234 hours of Tamasheq audio, while our contrastive model uses a French phonetic transcription of the Tamasheq audio as input in a Conformer speech translation architecture jointly trained on automatic speech recognition, ST and machine translation losses. Our results highlight that self-supervised models trained on smaller sets of target data are more effective to low-resource end-to-end ST fine-tuning, compared to large off-the-shelf models. Results also illustrate that even approximate phonetic transcriptions can improve ST scores.

Victoria Mingote, Pablo Gimeno, Luis Vicente et al. · mit

This paper proposes a direct text to speech translation system using discrete acoustic units. This framework employs text in different source languages as input to generate speech in the target language without the need for text transcriptions in this language. Motivated by the success of acoustic units in previous works for direct speech to speech translation systems, we use the same pipeline to extract the acoustic units using a speech encoder combined with a clustering algorithm. Once units are obtained, an encoder-decoder architecture is trained to predict them. Then a vocoder generates speech from units. Our approach for direct text to speech translation was tested on the new CVSS corpus with two different text mBART models employed as initialisation. The systems presented report competitive performance for most of the language pairs evaluated. Besides, results show a remarkable improvement when initialising our proposed architecture with a model pre-trained with more languages.

Valentin Pelloin, Franck Dary, Nicolas Herve et al.

We aim at improving spoken language modeling (LM) using very large amount of automatically transcribed speech. We leverage the INA (French National Audiovisual Institute) collection and obtain 19GB of text after applying ASR on 350,000 hours of diverse TV shows. From this, spoken language models are trained either by fine-tuning an existing LM (FlauBERT) or through training a LM from scratch. New models (FlauBERT-Oral) are shared with the community and evaluated for 3 downstream tasks: spoken language understanding, classification of TV shows and speech syntactic parsing. Results show that FlauBERT-Oral can be beneficial compared to its initial FlauBERT version demonstrating that, despite its inherent noisy nature, ASR-generated text can be used to build spoken language models.

Martin Lebourdais, Marie Tahon, Antoine Laurent et al.

This article focuses on overlapped speech and gender detection in order to study interactions between women and men in French audiovisual media (Gender Equality Monitoring project). In this application context, we need to automatically segment the speech signal according to speakers gender, and to identify when at least two speakers speak at the same time. We propose to use WavLM model which has the advantage of being pre-trained on a huge amount of speech data, to build an overlapped speech detection (OSD) and a gender detection (GD) systems. In this study, we use two different corpora. The DIHARD III corpus which is well adapted for the OSD task but lack gender information. The ALLIES corpus fits with the project application context. Our best OSD system is a Temporal Convolutional Network (TCN) with WavLM pre-trained features as input, which reaches a new state-of-the-art F1-score performance on DIHARD. A neural GD is trained with WavLM inputs on a gender balanced subset of the French broadcast news ALLIES data, and obtains an accuracy of 97.9%. This work opens new perspectives for human science researchers regarding the differences of representation between women and men in French media.

Martin Lebourdais, Théo Mariotte, Marie Tahon et al.

Voice activity and overlapped speech detection (respectively VAD and OSD) are key pre-processing tasks for speaker diarization. The final segmentation performance highly relies on the robustness of these sub-tasks. Recent studies have shown VAD and OSD can be trained jointly using a multi-class classification model. However, these works are often restricted to a specific speech domain, lacking information about the generalization capacities of the systems. This paper proposes a complete and new benchmark of different VAD and OSD models, on multiple audio setups (single/multi-channel) and speech domains (e.g. media, meeting...). Our 2/3-class systems, which combine a Temporal Convolutional Network with speech representations adapted to the setup, outperform state-of-the-art results. We show that the joint training of these two tasks offers similar performances in terms of F1-score to two dedicated VAD and OSD systems while reducing the training cost. This unique architecture can also be used for single and multichannel speech processing.

Mohammad Mohammadamini, Daban Q. Jaff, Josep Crego et al.

We present KUTED, a speech-to-text translation (S2TT) dataset for Central Kurdish, derived from TED and TEDx talks. The corpus comprises 91,000 sentence pairs, including 170 hours of English audio, 1.65 million English tokens, and 1.40 million Central Kurdish tokens. We evaluate KUTED on the S2TT task and find that orthographic variation significantly degrades Kurdish translation performance, producing nonstandard outputs. To address this, we propose a systematic text standardization approach that yields substantial performance gains and more consistent translations. On a test set separated from TED talks, a fine-tuned Seamless model achieves 15.18 BLEU, and we improve Seamless baseline by 3.0 BLEU on the FLEURS benchmark. We also train a Transformer model from scratch and evaluate a cascaded system that combines Seamless (ASR) with NLLB (MT).

Rémi Uro, David Doukhan, Albert Rilliard et al.

This paper presents a semi-automatic approach to create a diachronic corpus of voices balanced for speaker's age, gender, and recording period, according to 32 categories (2 genders, 4 age ranges and 4 recording periods). Corpora were selected at French National Institute of Audiovisual (INA) to obtain at least 30 speakers per category (a total of 960 speakers; only 874 have be found yet). For each speaker, speech excerpts were extracted from audiovisual documents using an automatic pipeline consisting of speech detection, background music and overlapped speech removal and speaker diarization, used to present clean speaker segments to human annotators identifying target speakers. This pipeline proved highly effective, cutting down manual processing by a factor of ten. Evaluation of the quality of the automatic processing and of the final output is provided. It shows the automatic processing compare to up-to-date process, and that the output provides high quality speech for most of the selected excerpts. This method shows promise for creating large corpora of known target speakers.

Sameer Khurana, Dominik Klement, Antoine Laurent et al.

We propose Hierarchical Audio Codec (HAC), a unified neural speech codec that factorizes its bottleneck into three linguistic levels-acoustic, phonetic, and lexical-within a single model. HAC leverages two knowledge distillation objectives: one from a pre-trained speech encoder (HuBERT) for phoneme-level structure, and another from a text-based encoder (LaBSE) for lexical cues. Experiments on English and multilingual data show that HAC's factorized bottleneck yields disentangled token sets: one aligns with phonemes, while another captures word-level semantics. Quantitative evaluations confirm that HAC tokens preserve naturalness and provide interpretable linguistic information, outperforming single-level baselines in both disentanglement and reconstruction quality. These findings underscore HAC's potential as a unified discrete speech representation, bridging acoustic detail and lexical meaning for downstream speech generation and understanding tasks.

Santiago Cuervo, Adel Moumen, Yanis Labrak et al. · mit

Text-Speech Language Models (TSLMs) -- language models trained to jointly process and generate text and speech -- are commonly trained through an early modality fusion/fission approach, in which both modalities are fed and predicted from a shared backbone via linear layers. We hypothesize that this approach limits cross-modal transfer by neglecting feature compositionality -- specifically, the finer-grained nature of speech representations compared to text -- preventing the emergence of a shared feature hierarchy within model layers. In this paper, we argue that this limitation can be addressed through late fusion and fission, with a fission process that accesses both high- and low-level features for speech generation. Our models implementing these principles, SmolTolk, rival or surpass state-of-the-art TSLMs trained with orders of magnitude more compute, and achieve significantly improved cross-modal performance relative to early fusion/fission baselines. Representation analyses further suggest that our method enhances the model's ability to abstract higher-level, more semantic features from speech, and leads to increasingly shared representation spaces across layers.

Antoine Laurent, Jean Mélou, Catherine Schwab et al.

The need to digitize heritage objects is now widely accepted. This article presents the very fashionable context of the creation of ''digital twins''. It illustrates the diversity of photographic 3D digitization methods, but this is not its only objective. Using a selection of objects from the collections of the mus{é}e d'Arch{é}ologie nationale, it shows that no single method is suitable for all cases. Rather, the method to be recommended for a given object should be the result of a concerted choice between those involved in heritage and those involved in the digital domain, as each new object may require the adaptation of existing tools. It would therefore be pointless to attempt an absolute classification of 3D digitization methods. On the contrary, we need to find the digital tool best suited to each object, taking into account not only its characteristics, but also the future use of its digital twin.

Rémi Uro, Marie Tahon, David Doukhan et al.

Transition Relevance Places are defined as the end of an utterance where the interlocutor may take the floor without interrupting the current speaker --i.e., a place where the turn is terminal. Analyzing turn terminality is useful to study the dynamic of turn-taking in spontaneous conversations. This paper presents an automatic classification of spoken utterances as Terminal or Non-Terminal in multi-speaker settings. We compared audio, text, and fusions of both approaches on a French corpus of TV and Radio extracts annotated with turn-terminality information at each speaker change. Our models are based on pre-trained self-supervised representations. We report results for different fusion strategies and varying context sizes. This study also questions the problem of performance variability by analyzing the differences in results for multiple training runs with random initialization. The measured accuracy would allow the use of these models for large-scale analysis of turn-taking.

Sameer Khurana, Antoine Laurent, James Glass

We propose a simple and effective cross-lingual transfer learning method to adapt monolingual wav2vec-2.0 models for Automatic Speech Recognition (ASR) in resource-scarce languages. We show that a monolingual wav2vec-2.0 is a good few-shot ASR learner in several languages. We improve its performance further via several iterations of Dropout Uncertainty-Driven Self-Training (DUST) by using a moderate-sized unlabeled speech dataset in the target language. A key finding of this work is that the adapted monolingual wav2vec-2.0 achieves similar performance as the topline multilingual XLSR model, which is trained on fifty-three languages, on the target language ASR task.

Hervé Bredin, Antoine Laurent

Speaker segmentation consists in partitioning a conversation between one or more speakers into speaker turns. Usually addressed as the late combination of three sub-tasks (voice activity detection, speaker change detection, and overlapped speech detection), we propose to train an end-to-end segmentation model that does it directly. Inspired by the original end-to-end neural speaker diarization approach (EEND), the task is modeled as a multi-label classification problem using permutation-invariant training. The main difference is that our model operates on short audio chunks (5 seconds) but at a much higher temporal resolution (every 16ms). Experiments on multiple speaker diarization datasets conclude that our model can be used with great success on both voice activity detection and overlapped speech detection. Our proposed model can also be used as a post-processing step, to detect and correctly assign overlapped speech regions. Relative diarization error rate improvement over the best considered baseline (VBx) reaches 17% on AMI, 13% on DIHARD 3, and 13% on VoxConverse.

Valentin Pelloin, Nathalie Camelin, Antoine Laurent et al.

In this paper, we propose a novel end-to-end sequence-to-sequence spoken language understanding model using an attention mechanism. It reliably selects contextual acoustic features in order to hypothesize semantic contents. An initial architecture capable of extracting all pronounced words and concepts from acoustic spans is designed and tested. With a shallow fusion language model, this system reaches a 13.6 concept error rate (CER) and an 18.5 concept value error rate (CVER) on the French MEDIA corpus, achieving an absolute 2.8 points reduction compared to the state-of-the-art. Then, an original model is proposed for hypothesizing concepts and their values. This transduction reaches a 15.4 CER and a 21.6 CVER without any new type of context.

Sameer Khurana, Antoine Laurent, James Glass

More than half of the 7,000 languages in the world are in imminent danger of going extinct. Traditional methods of documenting language proceed by collecting audio data followed by manual annotation by trained linguists at different levels of granularity. This time consuming and painstaking process could benefit from machine learning. Many endangered languages do not have any orthographic form but usually have speakers that are bi-lingual and trained in a high resource language. It is relatively easy to obtain textual translations corresponding to speech. In this work, we provide a multimodal machine learning framework for speech representation learning by exploiting the correlations between the two modalities namely speech and its corresponding text translation. Here, we construct a convolutional neural network audio encoder capable of extracting linguistic representations from speech. The audio encoder is trained to perform a speech-translation retrieval task in a contrastive learning framework. By evaluating the learned representations on a phone recognition task, we demonstrate that linguistic representations emerge in the audio encoder's internal representations as a by-product of learning to perform the retrieval task.

Sameer Khurana, Antoine Laurent, Wei-Ning Hsu et al.

Probabilistic Latent Variable Models (LVMs) provide an alternative to self-supervised learning approaches for linguistic representation learning from speech. LVMs admit an intuitive probabilistic interpretation where the latent structure shapes the information extracted from the signal. Even though LVMs have recently seen a renewed interest due to the introduction of Variational Autoencoders (VAEs), their use for speech representation learning remains largely unexplored. In this work, we propose Convolutional Deep Markov Model (ConvDMM), a Gaussian state-space model with non-linear emission and transition functions modelled by deep neural networks. This unsupervised model is trained using black box variational inference. A deep convolutional neural network is used as an inference network for structured variational approximation. When trained on a large scale speech dataset (LibriSpeech), ConvDMM produces features that significantly outperform multiple self-supervised feature extracting methods on linear phone classification and recognition on the Wall Street Journal dataset. Furthermore, we found that ConvDMM complements self-supervised methods like Wav2Vec and PASE, improving on the results achieved with any of the methods alone. Lastly, we find that ConvDMM features enable learning better phone recognizers than any other features in an extreme low-resource regime with few labeled training examples.

Adrian Łańcucki, Jan Chorowski, Guillaume Sanchez et al.

In this paper we demonstrate methods for reliable and efficient training of discrete representation using Vector-Quantized Variational Auto-Encoder models (VQ-VAEs). Discrete latent variable models have been shown to learn nontrivial representations of speech, applicable to unsupervised voice conversion and reaching state-of-the-art performance on unit discovery tasks. For unsupervised representation learning, they became viable alternatives to continuous latent variable models such as the Variational Auto-Encoder (VAE). However, training deep discrete variable models is challenging, due to the inherent non-differentiability of the discretization operation. In this paper we focus on VQ-VAE, a state-of-the-art discrete bottleneck model shown to perform on par with its continuous counterparts. It quantizes encoder outputs with on-line $k$-means clustering. We show that the codebook learning can suffer from poor initialization and non-stationarity of clustered encoder outputs. We demonstrate that these can be successfully overcome by increasing the learning rate for the codebook and periodic date-dependent codeword re-initialization. As a result, we achieve more robust training across different tasks, and significantly increase the usage of latent codewords even for large codebooks. This has practical benefit, for instance, in unsupervised representation learning, where large codebooks may lead to disentanglement of latent representations.

Natalia Tomashenko, Antoine Caubriere, Yannick Esteve et al.

This work investigates spoken language understanding (SLU) systems in the scenario when the semantic information is extracted directly from the speech signal by means of a single end-to-end neural network model. Two SLU tasks are considered: named entity recognition (NER) and semantic slot filling (SF). For these tasks, in order to improve the model performance, we explore various techniques including speaker adaptation, a modification of the connectionist temporal classification (CTC) training criterion, and sequential pretraining.

Antoine Caubrière, Natalia Tomashenko, Antoine Laurent et al.

We present an end-to-end approach to extract semantic concepts directly from the speech audio signal. To overcome the lack of data available for this spoken language understanding approach, we investigate the use of a transfer learning strategy based on the principles of curriculum learning. This approach allows us to exploit out-of-domain data that can help to prepare a fully neural architecture. Experiments are carried out on the French MEDIA and PORTMEDIA corpora and show that this end-to-end SLU approach reaches the best results ever published on this task. We compare our approach to a classical pipeline approach that uses ASR, POS tagging, lemmatizer, chunker... and other NLP tools that aim to enrich ASR outputs that feed an SLU text to concepts system. Last, we explore the promising capacity of our end-to-end SLU approach to address the problem of domain portability.

Sahar Ghannay, Antoine Caubrière, Yannick Estève et al.

Named entity recognition (NER) is among SLU tasks that usually extract semantic information from textual documents. Until now, NER from speech is made through a pipeline process that consists in processing first an automatic speech recognition (ASR) on the audio and then processing a NER on the ASR outputs. Such approach has some disadvantages (error propagation, metric to tune ASR systems sub-optimal in regards to the final task, reduced space search at the ASR output level...) and it is known that more integrated approaches outperform sequential ones, when they can be applied. In this paper, we present a first study of end-to-end approach that directly extracts named entities from speech, though a unique neural architecture. On a such way, a joint optimization is able for both ASR and NER. Experiments are carried on French data easily accessible, composed of data distributed in several evaluation campaign. Experimental results show that this end-to-end approach provides better results (F-measure=0.69 on test data) than a classical pipeline approach to detect named entity categories (F-measure=0.65).