Maureen de Seyssel

Maureen de Seyssel, Guillaume Wisniewski, Emmanuel Dupoux · apple-ml

According to the Language Familiarity Effect (LFE), people are better at discriminating between speakers of their native language. Although this cognitive effect was largely studied in the literature, experiments have only been conducted on a limited number of language pairs and their results only show the presence of the effect without yielding a gradual measure that may vary across language pairs. In this work, we show that the computational model of LFE introduced by Thorburn, Feldmand and Schatz (2019) can address these two limitations. In a first experiment, we attest to this model's capacity to obtain a gradual measure of the LFE by replicating behavioural findings on native and accented speech. In a second experiment, we evaluate LFE on a large number of language pairs, including many which have never been tested on humans. We show that the effect is replicated across a wide array of languages, providing further evidence of its universality. Building on the gradual measure of LFE, we also show that languages belonging to the same family yield smaller scores, supporting the idea of an effect of language distance on LFE.

Maureen de Seyssel, Marvin Lavechin, Hadrien Titeux et al. · apple-ml

We present ProsAudit, a benchmark in English to assess structural prosodic knowledge in self-supervised learning (SSL) speech models. It consists of two subtasks, their corresponding metrics, and an evaluation dataset. In the protosyntax task, the model must correctly identify strong versus weak prosodic boundaries. In the lexical task, the model needs to correctly distinguish between pauses inserted between words and within words. We also provide human evaluation scores on this benchmark. We evaluated a series of SSL models and found that they were all able to perform above chance on both tasks, even when evaluated on an unseen language. However, non-native models performed significantly worse than native ones on the lexical task, highlighting the importance of lexical knowledge in this task. We also found a clear effect of size with models trained on more data performing better in the two subtasks.

Tu Anh Nguyen, Maureen de Seyssel, Robin Algayres et al. · apple-ml

Word or word-fragment based Language Models (LM) are typically preferred over character-based ones in many downstream applications. This may not be surprising as words seem more linguistically relevant units than characters. Words provide at least two kinds of relevant information: boundary information and meaningful units. However, word boundary information may be absent or unreliable in the case of speech input (word boundaries are not marked explicitly in the speech stream). Here, we systematically compare LSTMs as a function of the input unit (character, phoneme, word, word part), with or without gold boundary information. We probe linguistic knowledge in the networks at the lexical, syntactic and semantic levels using three speech-adapted black box NLP psycholinguistically-inpired benchmarks (pWUGGY, pBLIMP, pSIMI). We find that the absence of boundaries costs between 2\% and 28\% in relative performance depending on the task. We show that gold boundaries can be replaced by automatically found ones obtained with an unsupervised segmentation algorithm, and that even modest segmentation performance gives a gain in performance on two of the three tasks compared to basic character/phone based models without boundary information.

Zakaria Aldeneh, Skyler Seto, Maureen de Seyssel et al.

Modern ASR systems are typically trained on large-scale pseudo-labeled, in-the-wild data spanning multiple domains. While such heterogeneous data benefit generalist models designed for broad deployment, they pose challenges for specialist models targeting specific domains: specialist models lack the capacity to learn from all available data, and one must pay closer attention to addressing the mismatch between training and test conditions. In this work, we study targeted data selection as a strategy to address these challenges, selecting relevant subsets from 100k hours of in-the-wild training data to optimize performance on target domains. We represent speech samples using embeddings that capture complementary characteristic--speaker attributes, phonetic content, and semantic meaning--and analyze how relevance and diversity along these axes when performing data selection affect downstream ASR performance. Our experiments with CTC-based Conformer models show that training on a strategically selected 5% subset can exceed the performance of models trained on the full dataset by up to 36.8% relative WER reduction on target domains.

Maureen de Seyssel, Antony D'Avirro, Adina Williams et al.

We introduce EmphAssess, a prosodic benchmark designed to evaluate the capability of speech-to-speech models to encode and reproduce prosodic emphasis. We apply this to two tasks: speech resynthesis and speech-to-speech translation. In both cases, the benchmark evaluates the ability of the model to encode emphasis in the speech input and accurately reproduce it in the output, potentially across a change of speaker and language. As part of the evaluation pipeline, we introduce EmphaClass, a new model that classifies emphasis at the frame or word level.

Jie Chi, Maureen de Seyssel, Natalie Schluter

Spoken language understanding research to date has generally carried a heavy text perspective. Most datasets are derived from text, which is then subsequently synthesized into speech, and most models typically rely on automatic transcriptions of speech. This is to the detriment of prosody--additional information carried by the speech signal beyond the phonetics of the words themselves and difficult to recover from text alone. In this work, we investigate the role of prosody in Spoken Question Answering. By isolating prosodic and lexical information on the SLUE-SQA-5 dataset, which consists of natural speech, we demonstrate that models trained on prosodic information alone can perform reasonably well by utilizing prosodic cues. However, we find that when lexical information is available, models tend to predominantly rely on it. Our findings suggest that while prosodic cues provide valuable supplementary information, more effective integration methods are required to ensure prosody contributes more significantly alongside lexical features.

Maureen de Seyssel, Eeshan Gunesh Dhekane · apple-ml

Speech foundation models have recently achieved remarkable capabilities across a wide range of tasks. However, their evaluation remains disjointed across tasks and model types. Different models excel at distinct aspects of speech processing and thus require different evaluation protocols. This paper proposes a unified taxonomy that addresses the question: Which evaluation is appropriate for which model? The taxonomy defines three orthogonal axes: the \textbf{evaluation aspect} being measured, the model capabilities required to attempt the task, and the task or protocol requirements needed to perform it. We classify a broad set of existing evaluations and benchmarks along these axes, spanning areas such as representation learning, speech generation, and interactive dialogue. By mapping each evaluation to the capabilities a model exposes (e.g., speech generation, real-time processing) and to its methodological demands (e.g., fine-tuning data, human judgment), the taxonomy provides a principled framework for aligning models with suitable evaluation methods. It also reveals systematic gaps, such as limited coverage of prosody, interaction, or reasoning, that highlight priorities for future benchmark design. Overall, this work offers a conceptual foundation and practical guide for selecting, interpreting, and extending evaluations of speech models.

Santiago Cuervo, Skyler Seto, Maureen de Seyssel et al.

Large Language Models (LLMs) can be adapted to extend their text capabilities to speech inputs. However, these speech-adapted LLMs consistently underperform their text-based counterparts--and even cascaded pipelines--on language understanding tasks. We term this shortfall the text-speech understanding gap: the performance drop observed when a speech-adapted LLM processes spoken inputs relative to when the original text-based LLM processes the equivalent text. Recent approaches to narrowing this gap either rely on large-scale speech synthesis of text corpora, which is costly and heavily dependent on synthetic data, or on large-scale proprietary speech datasets, which are not reproducible. As a result, there remains a need for more data-efficient alternatives for closing the text-speech understanding gap. In this work, we analyze the gap as driven by two factors: (i) forgetting of text capabilities during adaptation, and (ii) cross-modal misalignment between speech and text. Based on this analysis, we introduce SALAD--Sample-efficient Alignment with Learning through Active selection and cross-modal Distillation--which combines cross-modal distillation with targeted synthetic data to improve alignment while mitigating forgetting. Applied to 3B and 7B LLMs, SALAD achieves competitive performance with a strong open-weight model across broad-domain benchmarks in knowledge, language understanding, and reasoning, while training on over an order of magnitude less speech data from public corpora.

María Andrea Cruz Blandón, Zakaria Aldeneh, Jie Chi et al.

Self-supervised learning (SSL) has made significant advances in speech representation learning. Models like wav2vec 2.0 and HuBERT have achieved state-of-the-art results in tasks such as speech recognition, particularly in monolingual settings. However, multilingual SSL models tend to underperform their monolingual counterparts on each individual language, especially in multilingual scenarios with few languages such as the bilingual setting. In this work, we investigate a novel approach to reduce this performance gap by introducing limited visual grounding into bilingual speech SSL models. Our results show that visual grounding benefits both monolingual and bilingual models, with especially pronounced gains for the latter, reducing the multilingual performance gap on zero-shot phonetic discrimination from 31.5% for audio-only models to 8.04% with grounding.

Matthias Sperber, Maureen de Seyssel, Jiajun Bao et al.

Current speech translation systems, while having achieved impressive accuracies, are rather static in their behavior and do not adapt to real-world situations in ways human interpreters do. In order to improve their practical usefulness and enable interpreting-like experiences, a precise understanding of the nature of human interpreting is crucial. To this end, we discuss human interpreting literature from the perspective of the machine translation field, while considering both operational and qualitative aspects. We identify implications for the development of speech translation systems and argue that there is great potential to adopt many human interpreting principles using recent modeling techniques. We hope that our findings provide inspiration for closing the perceived usability gap, and can motivate progress toward true machine interpreting.

Skyler Seto, Maartje ter Hoeve, Maureen de Seyssel et al.

Bilingual and multilingual language models offer a promising path toward scaling NLP systems across diverse languages and users. However, their performance often varies wildly between languages as prior works show that adding more languages can degrade performance for some languages (such as English), while improving others (typically more data constrained languages). In this work, we investigate causes of these inconsistencies by comparing bilingual and monolingual language models. Our analysis reveals that unequal data quality, not just data quantity, is a major driver of performance degradation in bilingual settings. We propose a simple yet effective data filtering strategy to select higher-quality bilingual training data with only high quality English data. Applied to French, German, and Chinese, our approach improves monolingual performance by 2-4% and reduces bilingual model performance gaps to 1%. These results highlight the overlooked importance of data quality in multilingual pretraining and offer a practical recipe for balancing performance.

Maureen de Seyssel, Jie Chi, Skyler Seto et al. · apple-ml

We introduce a set of training-free ABX-style discrimination tasks to evaluate how multilingual language models represent language identity (form) and semantic content (meaning). Inspired from speech processing, these zero-shot tasks measure whether minimal differences in representation can be reliably detected. This offers a flexible and interpretable alternative to probing. Applied to XLM-R (Conneau et al, 2020) across pretraining checkpoints and layers, we find that language discrimination declines over training and becomes concentrated in lower layers, while meaning discrimination strengthens over time and stabilizes in deeper layers. We then explore probing tasks, showing some alignment between our metrics and linguistic learning performance. Our results position ABX tasks as a lightweight framework for analyzing the structure of multilingual representations.

Ewan Dunbar, Mathieu Bernard, Nicolas Hamilakis et al.

We present the Zero Resource Speech Challenge 2021, which asks participants to learn a language model directly from audio, without any text or labels. The challenge is based on the Libri-light dataset, which provides up to 60k hours of audio from English audio books without any associated text. We provide a pipeline baseline system consisting on an encoder based on contrastive predictive coding (CPC), a quantizer ($k$-means) and a standard language model (BERT or LSTM). The metrics evaluate the learned representations at the acoustic (ABX discrimination), lexical (spot-the-word), syntactic (acceptability judgment) and semantic levels (similarity judgment). We present an overview of the eight submitted systems from four groups and discuss the main results.

Tu Anh Nguyen, Maureen de Seyssel, Patricia Rozé et al.

We introduce a new unsupervised task, spoken language modeling: the learning of linguistic representations from raw audio signals without any labels, along with the Zero Resource Speech Benchmark 2021: a suite of 4 black-box, zero-shot metrics probing for the quality of the learned models at 4 linguistic levels: phonetics, lexicon, syntax and semantics. We present the results and analyses of a composite baseline made of the concatenation of three unsupervised systems: self-supervised contrastive representation learning (CPC), clustering (k-means) and language modeling (LSTM or BERT). The language models learn on the basis of the pseudo-text derived from clustering the learned representations. This simple pipeline shows better than chance performance on all four metrics, demonstrating the feasibility of spoken language modeling from raw speech. It also yields worse performance compared to text-based 'topline' systems trained on the same data, delineating the space to be explored by more sophisticated end-to-end models.

Estelle Maudet, Oralie Cattan, Maureen de Seyssel et al.

This paper reports on Qwant Research contribution to tasks 2 and 3 of the DEFT 2019's challenge, focusing on French clinical cases analysis. Task 2 is a task on semantic similarity between clinical cases and discussions. For this task, we propose an approach based on language models and evaluate the impact on the results of different preprocessings and matching techniques. For task 3, we have developed an information extraction system yielding very encouraging results accuracy-wise. We have experimented two different approaches, one based on the exclusive use of neural networks, the other based on a linguistic analysis.