We present Vakyansh, an end to end toolkit for Speech Recognition in Indic languages. India is home to almost 121 languages and around 125 crore speakers. Yet most of the languages are low resource in terms of data and pretrained models. Through Vakyansh, we introduce automatic data pipelines for data creation, model training, model evaluation and deployment. We create 14,000 hours of speech data in 23 Indic languages and train wav2vec 2.0 based pretrained models. These pretrained models are then finetuned to create state of the art speech recognition models for 18 Indic languages which are followed by language models and punctuation restoration models. We open source all these resources with a mission that this will inspire the speech community to develop speech first applications using our ASR models in Indic languages.
Accurate detection of hate speech against politicians, policy making and political ideas is crucial to maintain democracy and free speech. Unfortunately, the amount of labelled data necessary for training models to detect hate speech are limited and domain-dependent. In this paper, we address the issue of classification of hate speech against policy makers from Twitter in Italian, producing the first resource of this type in this language. We collected and annotated 1264 tweets, examined the cases of disagreements between annotators, and performed in-domain and cross-domain hate speech classifications with different features and algorithms. We achieved a performance of ROC AUC 0.83 and analyzed the most predictive attributes, also finding the different language features in the anti-policymakers and anti-immigration domains. Finally, we visualized networks of hashtags to capture the topics used in hateful and normal tweets.
Automatic inference of important paralinguistic information such as age from speech is an important area of research with numerous spoken language technology based applications. Speaker age estimation has applications in enabling personalization and age-appropriate curation of information and content. However, research in speaker age estimation in children is especially challenging due to paucity of relevant speech data representing the developmental spectrum, and the high signal variability especially intra age variability that complicates modeling. Most approaches in children speaker age estimation adopt methods directly from research on adult speech processing. In this paper, we propose features specific to children and focus on speaker's phone duration as an important biomarker of children's age. We propose phone duration modeling for predicting age from child's speech. To enable that, children speech is first forced aligned with the corresponding transcription to derive phone duration distributions. Statistical functionals are computed from phone duration distributions for each phoneme which are in turn used to train regression models to predict speaker age. Two children speech datasets are employed to demonstrate the robustness of phone duration features. We perform age regression experiments on age categories ranging from children studying in kindergarten to grade 10. Experimental results suggest phone durations contain important development-related information of children. Phonemes contributing most to estimation of children speaker age are analyzed and presented.
Recently, sequence-to-sequence (seq-to-seq) models have been successfully applied in text-to-speech (TTS) to synthesize speech for single-language text. To synthesize speech for multiple languages usually requires multi-lingual speech from the target speaker. However, it is both laborious and expensive to collect high-quality multi-lingual TTS data for the target speakers. In this paper, we proposed to use low-quality code-switched found data from the non-target speakers to achieve cross-lingual voice cloning for the target speakers. Experiments show that our proposed method can generate high-quality code-switched speech in the target voices in terms of both naturalness and speaker consistency. More importantly, we find that our method can achieve a comparable result to the state-of-the-art (SOTA) performance in cross-lingual voice cloning.
Most approaches to multi-talker overlapped speech separation and recognition assume that the number of simultaneously active speakers is given, but in realistic situations, it is typically unknown. To cope with this, we extend an iterative speech extraction system with mechanisms to count the number of sources and combine it with a single-talker speech recognizer to form the first end-to-end multi-talker automatic speech recognition system for an unknown number of active speakers. Our experiments show very promising performance in counting accuracy, source separation and speech recognition on simulated clean mixtures from WSJ0-2mix and WSJ0-3mix. Among others, we set a new state-of-the-art word error rate on the WSJ0-2mix database. Furthermore, our system generalizes well to a larger number of speakers than it ever saw during training, as shown in experiments with the WSJ0-4mix database.
This paper describes a novel design of a neural network-based speech generation model for learning prosodic representation.The problem of representation learning is formulated according to the information bottleneck (IB) principle. A modified VQ-VAE quantized layer is incorporated in the speech generation model to control the IB capacity and adjust the balance between reconstruction power and disentangle capability of the learned representation. The proposed model is able to learn word-level prosodic representations from speech data. With an optimized IB capacity, the learned representations not only are adequate to reconstruct the original speech but also can be used to transfer the prosody onto different textual content. Extensive results of the objective and subjective evaluation are presented to demonstrate the effect of IB capacity control, the effectiveness, and potential usage of the learned prosodic representation in controllable neural speech generation.
Sharing real-world speech utterances is key to the training and deployment of voice-based services. However, it also raises privacy risks as speech contains a wealth of personal data. Speaker anonymization aims to remove speaker information from a speech utterance while leaving its linguistic and prosodic attributes intact. State-of-the-art techniques operate by disentangling the speaker information (represented via a speaker embedding) from these attributes and re-synthesizing speech based on the speaker embedding of another speaker. Prior research in the privacy community has shown that anonymization often provides brittle privacy protection, even less so any provable guarantee. In this work, we show that disentanglement is indeed not perfect: linguistic and prosodic attributes still contain speaker information. We remove speaker information from these attributes by introducing differentially private feature extractors based on an autoencoder and an automatic speech recognizer, respectively, trained using noise layers. We plug these extractors in the state-of-the-art anonymization pipeline and generate, for the first time, differentially private utterances with a provable upper bound on the speaker information they contain. We evaluate empirically the privacy and utility resulting from our differentially private speaker anonymization approach on the LibriSpeech data set. Experimental results show that the generated utterances retain very high utility for automatic speech recognition training and inference, while being much better protected against strong adversaries who leverage the full knowledge of the anonymization process to try to infer the speaker identity.
Human speech data comprises a rich set of domain factors such as accent, syntactic and semantic variety, or acoustic environment. Previous work explores the effect of domain mismatch in automatic speech recognition between pre-training and fine-tuning as a whole but does not dissect the contribution of individual factors. In this paper, we present a controlled study to better understand the effect of such factors on the performance of pre-trained representations. To do so, we pre-train models either on modified natural speech or synthesized audio, with a single domain factor modified, and then measure performance on automatic speech recognition after fine tuning. Results show that phonetic domain factors play an important role during pre-training while grammatical and syntactic factors are far less important. To our knowledge, this is the first study to better understand the domain characteristics in self-supervised pre-training for speech.
Decades of research has studied how language learning infants learn to discriminate speech sounds, segment words, and associate words with their meanings. While gradual development of such capabilities is unquestionable, the exact nature of these skills and the underlying mental representations yet remains unclear. In parallel, computational studies have shown that basic comprehension of speech can be achieved by statistical learning between speech and concurrent referentially ambiguous visual input. These models can operate without prior linguistic knowledge such as representations of linguistic units, and without learning mechanisms specifically targeted at such units. This has raised the question of to what extent knowledge of linguistic units, such as phone(me)s, syllables, and words, could actually emerge as latent representations supporting the translation between speech and representations in other modalities, and without the units being proximal learning targets for the learner. In this study, we formulate this idea as the so-called latent language hypothesis (LLH), connecting linguistic representation learning to general predictive processing within and across sensory modalities. We review the extent that the audiovisual aspect of LLH is supported by the existing computational studies. We then explore LLH further in extensive learning simulations with different neural network models for audiovisual cross-situational learning, and comparing learning from both synthetic and real speech data. We investigate whether the latent representations learned by the networks reflect phonetic, syllabic, or lexical structure of input speech by utilizing an array of complementary evaluation metrics related to linguistic selectivity and temporal characteristics of the representations. As a result, we find that representations associated...
Neural networks based vocoders have recently demonstrated the powerful ability to synthesize high quality speech. These models usually generate samples by conditioning on some spectrum features, such as Mel-spectrum. However, these features are extracted by using speech analysis module including some processing based on the human knowledge. In this work, we proposed RawNet, a truly end-to-end neural vocoder, which use a coder network to learn the higher representation of signal, and an autoregressive voder network to generate speech sample by sample. The coder and voder together act like an auto-encoder network, and could be jointly trained directly on raw waveform without any human-designed features. The experiments on the Copy-Synthesis tasks show that RawNet can achieve the comparative synthesized speech quality with LPCNet, with a smaller model architecture and faster speech generation at the inference step.