Consider a multichannel Ambisonic recording containing a mixture of several reverberant speech signals. Retreiving the reverberant Ambisonic signals corresponding to the individual speech sources blindly from the mixture is a challenging task as it requires to estimate multiple signal channels for each source. In this work, we propose AmbiSep, a deep neural network-based plane-wave domain masking approach to solve this task. The masking network uses learned feature representations and transformers in a triple-path processing configuration. We train and evaluate the proposed network architecture on a spatialized WSJ0-2mix dataset, and show that the method achieves a multichannel scale-invariant signal-to-distortion ratio improvement of 17.7 dB on the blind test set, while preserving the spatial characteristics of the separated sounds.
We present an articulatory synthesis framework for the synthesis and manipulation of oral cancer speech for clinical decision making and alleviation of patient stress. Objective and subjective evaluations demonstrate that the framework has acceptable naturalness and is worth further investigation. A subsequent subjective vowel and consonant identification experiment showed that the articulatory synthesis system can manipulate the articulatory trajectories so that the synthesised speech reproduces problems present in the ground truth oral cancer speech.
Text Simplification is an ongoing problem in Natural Language Processing, solution to which has varied implications. In conjunction with the TSAR-2022 Workshop @EMNLP2022 Lexical Simplification is the process of reducing the lexical complexity of a text by replacing difficult words with easier to read (or understand) expressions while preserving the original information and meaning. This paper explains the work done by our team "teamPN" for English sub task. We created a modular pipeline which combines modern day transformers based models with traditional NLP methods like paraphrasing and verb sense disambiguation. We created a multi level and modular pipeline where the target text is treated according to its semantics(Part of Speech Tag). Pipeline is multi level as we utilize multiple source models to find potential candidates for replacement, It is modular as we can switch the source models and their weight-age in the final re-ranking.
Transcription of legal proceedings is very important to enable access to justice. However, speech transcription is an expensive and slow process. In this paper we describe part of a combined research and industrial project for building an automated transcription tool designed specifically for the Justice sector in the UK. We explain the challenges involved in transcribing court room hearings and the Natural Language Processing (NLP) techniques we employ to tackle these challenges. We will show that fine-tuning a generic off-the-shelf pre-trained Automatic Speech Recognition (ASR) system with an in-domain language model as well as infusing common phrases extracted with a collocation detection model can improve not only the Word Error Rate (WER) of the transcribed hearings but avoid critical errors that are specific of the legal jargon and terminology commonly used in British courts.
Sequence-to-Sequence (S2S) models have achieved remarkable success on various text generation tasks. However, learning complex structures with S2S models remains challenging as external neural modules and additional lexicons are often supplemented to predict non-textual outputs. We present a systematic study of S2S modeling using contained decoding on four core tasks: part-of-speech tagging, named entity recognition, constituency and dependency parsing, to develop efficient exploitation methods costing zero extra parameters. In particular, 3 lexically diverse linearization schemas and corresponding constrained decoding methods are designed and evaluated. Experiments show that although more lexicalized schemas yield longer output sequences that require heavier training, their sequences being closer to natural language makes them easier to learn. Moreover, S2S models using our constrained decoding outperform other S2S approaches using external resources. Our best models perform better than or comparably to the state-of-the-art for all 4 tasks, lighting a promise for S2S models to generate non-sequential structures.
Acoustic-to-articulatory inversion (AAI) is to obtain the movement of articulators from speech signals. Until now, achieving a speaker-independent AAI remains a challenge given the limited data. Besides, most current works only use audio speech as input, causing an inevitable performance bottleneck. To solve these problems, firstly, we pre-train a speech decomposition network to decompose audio speech into speaker embedding and content embedding as the new personalized speech features to adapt to the speaker-independent case. Secondly, to further improve the AAI, we propose a novel auxiliary feature network to estimate the lip auxiliary features from the above personalized speech features. Experimental results on three public datasets show that, compared with the state-of-the-art only using the audio speech feature, the proposed method reduces the average RMSE by 0.25 and increases the average correlation coefficient by 2.0% in the speaker-dependent case. More importantly, the average RMSE decreases by 0.29 and the average correlation coefficient increases by 5.0% in the speaker-independent case.
The spread of hatred that was formerly limited to verbal communications has rapidly moved over the Internet. Social media and community forums that allow people to discuss and express their opinions are becoming platforms for the spreading of hate messages. Many countries have developed laws to avoid online hate speech. They hold the companies that run the social media responsible for their failure to eliminate hate speech. But as online content continues to grow, so does the spread of hate speech However, manual analysis of hate speech on online platforms is infeasible due to the huge amount of data as it is expensive and time consuming. Thus, it is important to automatically process the online user contents to detect and remove hate speech from online media. Many recent approaches suffer from interpretability problem which means that it can be difficult to understand why the systems make the decisions they do. Through this work, some solutions for the problem of automatic detection of hate messages were proposed using Support Vector Machine (SVM) and Na\"ive Bayes algorithms. This achieved near state-of-the-art performance while being simpler and producing more easily interpretable decisions than other methods. Empirical evaluation of this technique has resulted in a classification accuracy of approximately 99% and 50% for SVM and NB respectively over the test set. Keywords: classification; hate speech; feature extraction, algorithm, supervised learning
Denoising diffusion probabilistic models (DDPMs) have recently achieved leading performances in many generative tasks. However, the inherited iterative sampling process costs hinder their applications to text-to-speech deployment. Through the preliminary study on diffusion model parameterization, we find that previous gradient-based TTS models require hundreds or thousands of iterations to guarantee high sample quality, which poses a challenge for accelerating sampling. In this work, we propose ProDiff, on progressive fast diffusion model for high-quality text-to-speech. Unlike previous work estimating the gradient for data density, ProDiff parameterizes the denoising model by directly predicting clean data to avoid distinct quality degradation in accelerating sampling. To tackle the model convergence challenge with decreased diffusion iterations, ProDiff reduces the data variance in the target site via knowledge distillation. Specifically, the denoising model uses the generated mel-spectrogram from an N-step DDIM teacher as the training target and distills the behavior into a new model with N/2 steps. As such, it allows the TTS model to make sharp predictions and further reduces the sampling time by orders of magnitude. Our evaluation demonstrates that ProDiff needs only 2 iterations to synthesize high-fidelity mel-spectrograms, while it maintains sample quality and diversity competitive with state-of-the-art models using hundreds of steps. ProDiff enables a sampling speed of 24x faster than real-time on a single NVIDIA 2080Ti GPU, making diffusion models practically applicable to text-to-speech synthesis deployment for the first time. Our extensive ablation studies demonstrate that each design in ProDiff is effective, and we further show that ProDiff can be easily extended to the multi-speaker setting. Audio samples are available at \url{https://ProDiff.github.io/.}
Self-supervised speech representation learning aims to extract meaningful factors from the speech signal that can later be used across different downstream tasks, such as speech and/or emotion recognition. Existing models, such as HuBERT, however, can be fairly large thus may not be suitable for edge speech applications. Moreover, realistic applications typically involve speech corrupted by noise and room reverberation, hence models need to provide representations that are robust to such environmental factors. In this study, we build on the so-called DistilHuBERT model, which distils HuBERT to a fraction of its original size, with three modifications, namely: (i) augment the training data with noise and reverberation, while the student model needs to distill the clean representations from the teacher model; (ii) introduce a curriculum learning approach where increasing levels of noise are introduced as the model trains, thus helping with convergence and with the creation of more robust representations; and (iii) introduce a multi-task learning approach where the model also reconstructs the clean waveform jointly with the distillation task, thus also acting as an enhancement step to ensure additional environment robustness to the representation. Experiments on three SUPERB tasks show the advantages of the proposed method not only relative to the original DistilHuBERT, but also to the original HuBERT, thus showing the advantages of the proposed method for ``in the wild'' edge speech applications.
Hate speech detection is a common downstream application of natural language processing (NLP) in the real world. In spite of the increasing accuracy, current data-driven approaches could easily learn biases from the imbalanced data distributions originating from humans. The deployment of biased models could further enhance the existing social biases. But unlike handling tabular data, defining and mitigating biases in text classifiers, which deal with unstructured data, are more challenging. A popular solution for improving machine learning fairness in NLP is to conduct the debiasing process with a list of potentially discriminated words given by human annotators. In addition to suffering from the risks of overlooking the biased terms, exhaustively identifying bias with human annotators are unsustainable since discrimination is variable among different datasets and may evolve over time. To this end, we propose an automatic misuse detector (MiD) relying on an explanation method for detecting potential bias. And built upon that, an end-to-end debiasing framework with the proposed staged correction is designed for text classifiers without any external resources required.