Multilingual Speech Recognition for Low-Resource Indian Languages using Multi-Task conformer

Transformers have recently become very popular for sequence-to-sequence applications such as machine translation and speech recognition. In this work, we propose a multi-task learning-based transformer model for low-resource multilingual speech recognition for Indian languages. Our proposed model consists of a conformer [1] encoder and two parallel transformer decoders. We use a phoneme decoder (PHN-DEC) for the phoneme recognition task and a grapheme decoder (GRP-DEC) to predict grapheme sequence. We consider the phoneme recognition task as an auxiliary task for our multi-task learning framework. We jointly optimize the network for both phoneme and grapheme recognition tasks using Joint CTC-Attention [2] training. We use a conditional decoding scheme to inject the language information into the model before predicting the grapheme sequence. Our experiments show that our proposed approach can obtain significant improvement over previous approaches [4]. We also show that our conformer-based dual-decoder approach outperforms both the transformer-based dual-decoder approach and single decoder approach. Finally, We compare monolingual ASR models with our proposed multilingual ASR approach.

HateXplain: A Benchmark Dataset for Explainable Hate Speech Detection

Hate speech is a challenging issue plaguing the online social media. While better models for hate speech detection are continuously being developed, there is little research on the bias and interpretability aspects of hate speech. In this paper, we introduce HateXplain, the first benchmark hate speech dataset covering multiple aspects of the issue. Each post in our dataset is annotated from three different perspectives: the basic, commonly used 3-class classification (i.e., hate, offensive or normal), the target community (i.e., the community that has been the victim of hate speech/offensive speech in the post), and the rationales, i.e., the portions of the post on which their labelling decision (as hate, offensive or normal) is based. We utilize existing state-of-the-art models and observe that even models that perform very well in classification do not score high on explainability metrics like model plausibility and faithfulness. We also observe that models, which utilize the human rationales for training, perform better in reducing unintended bias towards target communities. We have made our code and dataset public at https://github.com/punyajoy/HateXplain

MPC-Pipe: an Efficient Pipeline Scheme for Secure Multi-party Machine Learning Inference

Multi-party computing (MPC) has been gaining popularity over the past years as a secure computing model, particularly for machine learning (ML) inference. Compared with its competitors, MPC has fewer overheads than homomorphic encryption (HE) and has a more robust threat model than hardware-based trusted execution environments (TEE) such as Intel SGX. Despite its apparent advantages, MPC protocols still pay substantial performance penalties compared to plaintext when applied to ML algorithms. The overhead is due to added computation and communication costs. For multiplications that are ubiquitous in ML algorithms, MPC protocols add 32x more computational costs and 1 round of broadcasting among MPC servers. Moreover, ML computations that have trivial costs in plaintext, such as Softmax, ReLU, and other non-linear operations become very expensive due to added communication. Those added overheads make MPC less palatable to deploy in real-time ML inference frameworks, such as speech translation. In this work, we present MPC-Pipe, an MPC pipeline inference technique that uses two ML-specific approaches. 1) inter-linear-layer pipeline and 2) inner layer pipeline. Those two techniques shorten the total inference runtime for machine learning models. Our experiments have shown to reduce ML inference latency by up to 12.6% when model weights are private and 14.48\% when model weights are public, compared to current MPC protocol implementations.

Improving End-to-End Contextual Speech Recognition with Fine-grained Contextual Knowledge Selection

Nowadays, most methods in end-to-end contextual speech recognition bias the recognition process towards contextual knowledge. Since all-neural contextual biasing methods rely on phrase-level contextual modeling and attention-based relevance modeling, they may encounter confusion between similar context-specific phrases, which hurts predictions at the token level. In this work, we focus on mitigating confusion problems with fine-grained contextual knowledge selection (FineCoS). In FineCoS, we introduce fine-grained knowledge to reduce the uncertainty of token predictions. Specifically, we first apply phrase selection to narrow the range of phrase candidates, and then conduct token attention on the tokens in the selected phrase candidates. Moreover, we re-normalize the attention weights of most relevant phrases in inference to obtain more focused phrase-level contextual representations, and inject position information to better discriminate phrases or tokens. On LibriSpeech and an in-house 160,000-hour dataset, we explore the proposed methods based on a controllable all-neural biasing method, collaborative decoding (ColDec). The proposed methods provide at most 6.1% relative word error rate reduction on LibriSpeech and 16.4% relative character error rate reduction on the in-house dataset over ColDec.

Improving Confidence Estimation on Out-of-Domain Data for End-to-End Speech Recognition

As end-to-end automatic speech recognition (ASR) models reach promising performance, various downstream tasks rely on good confidence estimators for these systems. Recent research has shown that model-based confidence estimators have a significant advantage over using the output softmax probabilities. If the input data to the speech recogniser is from mismatched acoustic and linguistic conditions, the ASR performance and the corresponding confidence estimators may exhibit severe degradation. Since confidence models are often trained on the same in-domain data as the ASR, generalising to out-of-domain (OOD) scenarios is challenging. By keeping the ASR model untouched, this paper proposes two approaches to improve the model-based confidence estimators on OOD data: using pseudo transcriptions and an additional OOD language model. With an ASR model trained on LibriSpeech, experiments show that the proposed methods can significantly improve the confidence metrics on TED-LIUM and Switchboard datasets while preserving in-domain performance. Furthermore, the improved confidence estimators are better calibrated on OOD data and can provide a much more reliable criterion for data selection.

Enhancing Speech Intelligibility in Text-To-Speech Synthesis using Speaking Style Conversion

The increased adoption of digital assistants makes text-to-speech (TTS) synthesis systems an indispensable feature of modern mobile devices. It is hence desirable to build a system capable of generating highly intelligible speech in the presence of noise. Past studies have investigated style conversion in TTS synthesis, yet degraded synthesized quality often leads to worse intelligibility. To overcome such limitations, we proposed a novel transfer learning approach using Tacotron and WaveRNN based TTS synthesis. The proposed speech system exploits two modification strategies: (a) Lombard speaking style data and (b) Spectral Shaping and Dynamic Range Compression (SSDRC) which has been shown to provide high intelligibility gains by redistributing the signal energy on the time-frequency domain. We refer to this extension as Lombard-SSDRC TTS system. Intelligibility enhancement as quantified by the Intelligibility in Bits (SIIB-Gauss) measure shows that the proposed Lombard-SSDRC TTS system shows significant relative improvement between 110% and 130% in speech-shaped noise (SSN), and 47% to 140% in competing-speaker noise (CSN) against the state-of-the-art TTS approach. Additional subjective evaluation shows that Lombard-SSDRC TTS successfully increases the speech intelligibility with relative improvement of 455% for SSN and 104% for CSN in median keyword correction rate compared to the baseline TTS method.

The Vocal Signature of Social Anxiety: Exploration using Hypothesis-Testing and Machine-Learning Approaches

Background - Social anxiety (SA) is a common and debilitating condition, negatively affecting life quality even at sub-diagnostic thresholds. We sought to characterize SA's acoustic signature using hypothesis-testing and machine learning (ML) approaches. Methods - Participants formed spontaneous utterances responding to instructions to refuse or consent to commands of alleged peers. Vocal properties (e.g., intensity and duration) of these utterances were analyzed. Results - Our prediction that, as compared to low-SA (n=31), high-SA (n=32) individuals exhibit a less confident vocal speech signature, especially with respect to refusal utterances, was only partially supported by the classical hypothesis-testing approach. However, the results of the ML analyses and specifically the decision tree classifier were consistent with such speech patterns in SA. Using a Gaussian Process (GP) classifier, we were able to distinguish between high- and low-SA individuals with high (75.6%) accuracy and good (.83 AUC) separability. We also expected and found that vocal properties differentiated between refusal and consent utterances. Conclusions - Our findings provide further support for the usefulness of ML approach for the study of psychopathology, highlighting the utility of developing automatic techniques to create behavioral markers of SAD. Clinically, the simplicity and accessibility of these procedures may encourage people to seek professional help.

Dissecting User-Perceived Latency of On-Device E2E Speech Recognition

As speech-enabled devices such as smartphones and smart speakers become increasingly ubiquitous, there is growing interest in building automatic speech recognition (ASR) systems that can run directly on-device; end-to-end (E2E) speech recognition models such as recurrent neural network transducers and their variants have recently emerged as prime candidates for this task. Apart from being accurate and compact, such systems need to decode speech with low user-perceived latency (UPL), producing words as soon as they are spoken. This work examines the impact of various techniques -- model architectures, training criteria, decoding hyperparameters, and endpointer parameters -- on UPL. Our analyses suggest that measures of model size (parameters, input chunk sizes), or measures of computation (e.g., FLOPS, RTF) that reflect the model's ability to process input frames are not always strongly correlated with observed UPL. Thus, conventional algorithmic latency measurements might be inadequate in accurately capturing latency observed when models are deployed on embedded devices. Instead, we find that factors affecting token emission latency, and endpointing behavior significantly impact on UPL. We achieve the best trade-off between latency and word error rate when performing ASR jointly with endpointing, and using the recently proposed alignment regularization.

Few-Shot Cross-Lingual TTS Using Transferable Phoneme Embedding

This paper studies a transferable phoneme embedding framework that aims to deal with the cross-lingual text-to-speech (TTS) problem under the few-shot setting. Transfer learning is a common approach when it comes to few-shot learning since training from scratch on few-shot training data is bound to overfit. Still, we find that the naive transfer learning approach fails to adapt to unseen languages under extremely few-shot settings, where less than 8 minutes of data is provided. We deal with the problem by proposing a framework that consists of a phoneme-based TTS model and a codebook module to project phonemes from different languages into a learned latent space. Furthermore, by utilizing phoneme-level averaged self-supervised learned features, we effectively improve the quality of synthesized speeches. Experiments show that using 4 utterances, which is about 30 seconds of data, is enough to synthesize intelligible speech when adapting to an unseen language using our framework.

Adversarial Multi-Task Deep Learning for Noise-Robust Voice Activity Detection with Low Algorithmic Delay

Voice Activity Detection (VAD) is an important pre-processing step in a wide variety of speech processing systems. VAD should in a practical application be able to detect speech in both noisy and noise-free environments, while not introducing significant latency. In this work we propose using an adversarial multi-task learning method when training a supervised VAD. The method has been applied to the state-of-the-art VAD Waveform-based Voice Activity Detection. Additionally the performance of the VADis investigated under different algorithmic delays, which is an important factor in latency. Introducing adversarial multi-task learning to the model is observed to increase performance in terms of Area Under Curve (AUC), particularly in noisy environments, while the performance is not degraded at higher SNR levels. The adversarial multi-task learning is only applied in the training phase and thus introduces no additional cost in testing. Furthermore the correlation between performance and algorithmic delays is investigated, and it is observed that the VAD performance degradation is only moderate when lowering the algorithmic delay from 398 ms to 23 ms.