Marie Tahon

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.

Hubert Nourtel, Pierre Champion, Denis Jouvet et al.

Speech data carries a range of personal information, such as the speaker's identity and emotional state. These attributes can be used for malicious purposes. With the development of virtual assistants, a new generation of privacy threats has emerged. Current studies have addressed the topic of preserving speech privacy. One of them, the VoicePrivacy initiative aims to promote the development of privacy preservation tools for speech technology. The task selected for the VoicePrivacy 2020 Challenge (VPC) is about speaker anonymization. The goal is to hide the source speaker's identity while preserving the linguistic information. The baseline of the VPC makes use of a voice conversion. This paper studies the impact of the speaker anonymization baseline system of the VPC on emotional information present in speech utterances. Evaluation is performed following the VPC rules regarding the attackers' knowledge about the anonymization system. Our results show that the VPC baseline system does not suppress speakers' emotions against informed attackers. When comparing anonymized speech to original speech, the emotion recognition performance is degraded by 15\% relative to IEMOCAP data, similar to the degradation observed for automatic speech recognition used to evaluate the preservation of the linguistic information.

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).

Manon Macary, Marie Tahon, Yannick Estève et al.

The goal of our research is to automatically retrieve the satisfaction and the frustration in real-life call-center conversations. This study focuses an industrial application in which the customer satisfaction is continuously tracked down to improve customer services. To compensate the lack of large annotated emotional databases, we explore the use of pre-trained speech representations as a form of transfer learning towards AlloSat corpus. Moreover, several studies have pointed out that emotion can be detected not only in speech but also in facial trait, in biological response or in textual information. In the context of telephone conversations, we can break down the audio information into acoustic and linguistic by using the speech signal and its transcription. Our experiments confirms the large gain in performance obtained with the use of pre-trained features. Surprisingly, we found that the linguistic content is clearly the major contributor for the prediction of satisfaction and best generalizes to unseen data. Our experiments conclude to the definitive advantage of using CamemBERT representations, however the benefit of the fusion of acoustic and linguistic modalities is not as obvious. With models learnt on individual annotations, we found that fusion approaches are more robust to the subjectivity of the annotation task. This study also tackles the problem of performances variability and intends to estimate this variability from different views: weights initialization, confidence intervals and annotation subjectivity. A deep analysis on the linguistic content investigates interpretable factors able to explain the high contribution of the linguistic modality for this task.

Meysam Shamsi, Marie Tahon

There are two paradigms of emotion representation, categorical labeling and dimensional description in continuous space. Therefore, the emotion recognition task can be treated as a classification or regression. The main aim of this study is to investigate the relation between these two representations and propose a classification pipeline that uses only dimensional annotation. The proposed approach contains a regressor model which is trained to predict a vector of continuous values in dimensional representation for given speech audio. The output of this model can be interpreted as an emotional category using a mapping algorithm. We investigated the performances of a combination of three feature extractors, three neural network architectures, and three mapping algorithms on two different corpora. Our study shows the advantages and limitations of the classification via regression approach.

Thibault Gaudier, Marie Tahon, Anthony Larcher et al.

Speech resynthesis is a generic task for which we want to synthesize audio with another audio as input, which finds applications for media monitors and journalists.Among different tasks addressed by speech resynthesis, voice conversion preserves the linguistic information while modifying the identity of the speaker, and speech edition preserves the identity of the speaker but some words are modified.In both cases, we need to disentangle speaker and phonetic contents in intermediate representations.Phonetic PosteriorGrams (PPG) are a frame-level probabilistic representation of phonemes, and are usually considered speaker-independent.This paper presents a PPG-based speech resynthesis system.A perceptive evaluation assesses that it produces correct audio quality.Then, we demonstrate that an automatic speaker verification model is not able to recover the source speaker after re-synthesis with PPG, even when the model is trained on synthetic data.

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.

Antonio Almudévar, Théo Mariotte, Alfonso Ortega et al.

Unsupervised Multiple Domain Translation is the task of transforming data from one domain to other domains without having paired data to train the systems. Typically, methods based on Generative Adversarial Networks (GANs) are used to address this task. However, our proposal exclusively relies on a modified version of a Variational Autoencoder. This modification consists of the use of two latent variables disentangled in a controlled way by design. One of this latent variables is imposed to depend exclusively on the domain, while the other one must depend on the rest of the variability factors of the data. Additionally, the conditions imposed over the domain latent variable allow for better control and understanding of the latent space. We empirically demonstrate that our approach works on different vision datasets improving the performance of other well known methods. Finally, we prove that, indeed, one of the latent variables stores all the information related to the domain and the other one hardly contains any domain information.

Antoine Teissier, Marie Tahon, Nicolas Dugué et al.

Due to the rapid progress of speech synthesis, deepfake detection has become a major concern in the speech processing community. Because it is a critical task, systems must not only be efficient and robust, but also provide interpretable explanations. Among the different approaches for explainability, we focus on the interpretation of latent representations. In such paper, we focus on the last layer of embeddings of AASIST, a deepfake detection architecture. We use a TopK activation inspired by SAEs on this layer to obtain sparse representations which are used in the decision process. We demonstrate that sparse deepfake detection can improve detection performance, with an EER of 23.36% on ASVSpoof5 test set, with 95% of sparsity. We then show that these representations provide better disentanglement, using completeness and modularity metrics based on mutual information. Notably, some attacks are directly encoded in the latent space.

Théo Mariotte, Martin Lebourdais, Antonio Almudévar et al.

Audio pretrained models are widely employed to solve various tasks in speech processing, sound event detection, or music information retrieval. However, the representations learned by these models are unclear, and their analysis mainly restricts to linear probing of the hidden representations. In this work, we explore the use of Sparse Autoencoders (SAEs) to analyze the hidden representations of pretrained models, focusing on a case study in singing technique classification. We first demonstrate that SAEs retain both information about the original representations and class labels, enabling their internal structure to provide insights into self-supervised learning systems. Furthermore, we show that SAEs enhance the disentanglement of vocal attributes, establishing them as an effective tool for identifying the underlying factors encoded in the representations.

Antonio Almudévar, Théo Mariotte, Alfonso Ortega et al.

Prototypical Learning is based on the idea that there is a point (which we call prototype) around which the embeddings of a class are clustered. It has shown promising results in scenarios with little labeled data or to design explainable models. Typically, prototypes are either defined as the average of the embeddings of a class or are designed to be trainable. In this work, we propose to predefine prototypes following human-specified criteria, which simplify the training pipeline and brings different advantages. Specifically, in this work we explore two of these advantages: increasing the inter-class separability of embeddings and disentangling embeddings with respect to different variance factors, which can translate into the possibility of having explainable predictions. Finally, we propose different experiments that help to understand our proposal and demonstrate empirically the mentioned advantages.

Martin Lebourdais, Théo Mariotte, Antonio Almudévar et al.

Audio segmentation is a key task for many speech technologies, most of which are based on neural networks, usually considered as black boxes, with high-level performances. However, in many domains, among which health or forensics, there is not only a need for good performance but also for explanations about the output decision. Explanations derived directly from latent representations need to satisfy "good" properties, such as informativeness, compactness, or modularity, to be interpretable. In this article, we propose an explainable-by-design audio segmentation model based on non-negative matrix factorization (NMF) which is a good candidate for the design of interpretable representations. This paper shows that our model reaches good segmentation performances, and presents deep analyses of the latent representation extracted from the non-negative matrix. The proposed approach opens new perspectives toward the evaluation of interpretable representations according to "good" properties.

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.

Théo Mariotte, Antonio Almudévar, Marie Tahon et al.

Audio signal segmentation is a key task for automatic audio indexing. It consists of detecting the boundaries of class-homogeneous segments in the signal. In many applications, explainable AI is a vital process for transparency of decision-making with machine learning. In this paper, we propose an explainable multilabel segmentation model that solves speech activity (SAD), music (MD), noise (ND), and overlapped speech detection (OSD) simultaneously. This proxy uses the non-negative matrix factorization (NMF) to map the embedding used for the segmentation to the frequency domain. Experiments conducted on two datasets show similar performances as the pre-trained black box model while showing strong explainability features. Specifically, the frequency bins used for the decision can be easily identified at both the segment level (local explanations) and global level (class prototypes).

Manon Macary, Marie Tahon, Yannick Estève et al.

Pre-training for feature extraction is an increasingly studied approach to get better continuous representations of audio and text content. In the present work, we use wav2vec and camemBERT as self-supervised learned models to represent our data in order to perform continuous emotion recognition from speech (SER) on AlloSat, a large French emotional database describing the satisfaction dimension, and on the state of the art corpus SEWA focusing on valence, arousal and liking dimensions. To the authors' knowledge, this paper presents the first study showing that the joint use of wav2vec and BERT-like pre-trained features is very relevant to deal with continuous SER task, usually characterized by a small amount of labeled training data. Evaluated by the well-known concordance correlation coefficient (CCC), our experiments show that we can reach a CCC value of 0.825 instead of 0.592 when using MFCC in conjunction with word2vec word embedding on the AlloSat dataset.