Connectionist Temporal Classification has recently attracted a lot of interest as it offers an elegant approach to building acoustic models (AMs) for speech recognition. The CTC loss function maps an input sequence of observable feature vectors to an output sequence of symbols. Output symbols are conditionally independent of each other under CTC loss, so a language model (LM) can be incorporated conveniently during decoding, retaining the traditional separation of acoustic and linguistic components in ASR. For fixed vocabularies, Weighted Finite State Transducers provide a strong baseline for efficient integration of CTC AMs with n-gram LMs. Character-based neural LMs provide a straight forward solution for open vocabulary speech recognition and all-neural models, and can be decoded with beam search. Finally, sequence-to-sequence models can be used to translate a sequence of individual sounds into a word string. We compare the performance of these three approaches, and analyze their error patterns, which provides insightful guidance for future research and development in this important area.
The goal of this work is to recognise phrases and sentences being spoken by a talking face, with or without the audio. Unlike previous works that have focussed on recognising a limited number of words or phrases, we tackle lip reading as an open-world problem - unconstrained natural language sentences, and in the wild videos. Our key contributions are: (1) a 'Watch, Listen, Attend and Spell' (WLAS) network that learns to transcribe videos of mouth motion to characters; (2) a curriculum learning strategy to accelerate training and to reduce overfitting; (3) a 'Lip Reading Sentences' (LRS) dataset for visual speech recognition, consisting of over 100,000 natural sentences from British television. The WLAS model trained on the LRS dataset surpasses the performance of all previous work on standard lip reading benchmark datasets, often by a significant margin. This lip reading performance beats a professional lip reader on videos from BBC television, and we also demonstrate that visual information helps to improve speech recognition performance even when the audio is available.
The Gumbel-softmax distribution, or Concrete distribution, is often used to relax the discrete characteristics of a categorical distribution and enable back-propagation through differentiable reparameterization. Although it reliably yields low variance gradients, it still relies on a stochastic sampling process for optimization. In this work, we present a relaxed categorical analytic bound (ReCAB), a novel divergence-like metric which corresponds to the upper bound of the Kullback-Leibler divergence (KLD) of a relaxed categorical distribution. The proposed metric is easy to implement because it has a closed form solution, and empirical results show that it is close to the actual KLD. Along with this new metric, we propose a relaxed categorical analytic bound variational autoencoder (ReCAB-VAE) that successfully models both continuous and relaxed discrete latent representations. We implement an emotional text-to-speech synthesis system based on the proposed framework, and show that the proposed system flexibly and stably controls emotion expressions with better speech quality compared to baselines that use stochastic estimation or categorical distribution approximation.
This paper presents the details of our system designed for the Task 1 of Multimodal Information Based Speech Processing (MISP) Challenge 2021. The purpose of Task 1 is to leverage both audio and video information to improve the environmental robustness of far-field wake word spotting. In the proposed system, firstly, we take advantage of speech enhancement algorithms such as beamforming and weighted prediction error (WPE) to address the multi-microphone conversational audio. Secondly, several data augmentation techniques are applied to simulate a more realistic far-field scenario. For the video information, the provided region of interest (ROI) is used to obtain visual representation. Then the multi-layer CNN is proposed to learn audio and visual representations, and these representations are fed into our two-branch attention-based network which can be employed for fusion, such as transformer and conformed. The focal loss is used to fine-tune the model and improve the performance significantly. Finally, multiple trained models are integrated by casting vote to achieve our final 0.091 score.
End-to-end automatic speech recognition (ASR) directly maps input speech to a character sequence without using pronunciation lexica. However, in languages with thousands of characters, such as Japanese and Mandarin, modeling all these characters is problematic due to data scarcity. To alleviate the problem, we propose a multi-task learning model with explicit interaction between characters and syllables by utilizing Self-conditioned connectionist temporal classification (CTC) technique. While the original Self-conditioned CTC estimates character-level intermediate predictions by applying auxiliary CTC losses to a set of intermediate layers, the proposed method additionally estimates syllable-level intermediate predictions in another set of intermediate layers. The character-level and syllable-level predictions are alternately used as conditioning features to deal with mutual dependency between characters and syllables. Experimental results on Japanese and Mandarin datasets show that the proposed multi-sequence intermediate conditioning outperformed the conventional multi-task-based and Self-conditioned CTC-based methods.
We present a new approach to disentangle speaker voice and phone content by introducing new components to the VQ-VAE architecture for speech synthesis. The original VQ-VAE does not generalize well to unseen speakers or content. To alleviate this problem, we have incorporated a speaker encoder and speaker VQ codebook that learns global speaker characteristics entirely separate from the existing sub-phone codebooks. We also compare two training methods: self-supervised with global conditions and semi-supervised with speaker labels. Adding a speaker VQ component improves objective measures of speech synthesis quality (estimated MOS, speaker similarity, ASR-based intelligibility) and provides learned representations that are meaningful. Our speaker VQ codebook indices can be used in a simple speaker diarization task and perform slightly better than an x-vector baseline. Additionally, phones can be recognized from sub-phone VQ codebook indices in our semi-supervised VQ-VAE better than self-supervised with global conditions.
Today's cloud service architectures follow a "one size fits all" deployment strategy where the same service version instantiation is provided to the end users. However, consumers are broad and different applications have different accuracy and responsiveness requirements, which as we demonstrate renders the "one size fits all" approach inefficient in practice. We use a production-grade speech recognition engine, which serves several thousands of users, and an open source computer vision based system, to explain our point. To overcome the limitations of the "one size fits all" approach, we recommend Tolerance Tiers where each MLaaS tier exposes an accuracy/responsiveness characteristic, and consumers can programmatically select a tier. We evaluate our proposal on the CPU-based automatic speech recognition (ASR) engine and cutting-edge neural networks for image classification deployed on both CPUs and GPUs. The results show that our proposed approach provides an MLaaS cloud service architecture that can be tuned by the end API user or consumer to outperform the conventional "one size fits all" approach.
The recent success of transformer networks for neural machine translation and other NLP tasks has led to a surge in research work trying to apply it for speech recognition. Recent efforts studied key research questions around ways of combining positional embedding with speech features, and stability of optimization for large scale learning of transformer networks. In this paper, we propose replacing the sinusoidal positional embedding for transformers with convolutionally learned input representations. These contextual representations provide subsequent transformer blocks with relative positional information needed for discovering long-range relationships between local concepts. The proposed system has favorable optimization characteristics where our reported results are produced with fixed learning rate of 1.0 and no warmup steps. The proposed model reduces the word error rate (WER) by 12% and 16% relative to previously published work on Librispeech "dev other" and "test other" subsets respectively, when no extra LM text is provided. Full code to reproduce our results will be available online at the time of publication.
This paper describes the Arabic Multi-Genre Broadcast (MGB-2) Challenge for SLT-2016. Unlike last year's English MGB Challenge, which focused on recognition of diverse TV genres, this year, the challenge has an emphasis on handling the diversity in dialect in Arabic speech. Audio data comes from 19 distinct programmes from the Aljazeera Arabic TV channel between March 2005 and December 2015. Programmes are split into three groups: conversations, interviews, and reports. A total of 1,200 hours have been released with lightly supervised transcriptions for the acoustic modelling. For language modelling, we made available over 110M words crawled from Aljazeera Arabic website Aljazeera.net for a 10 year duration 2000-2011. Two lexicons have been provided, one phoneme based and one grapheme based. Finally, two tasks were proposed for this year's challenge: standard speech transcription, and word alignment. This paper describes the task data and evaluation process used in the MGB challenge, and summarises the results obtained.
In recent work we have presented a formal framework for linguistic annotation based on labeled acyclic digraphs. These `annotation graphs' offer a simple yet powerful method for representing complex annotation structures incorporating hierarchy and overlap. Here, we motivate and illustrate our approach using discourse-level annotations of text and speech data drawn from the CALLHOME, COCONUT, MUC-7, DAMSL and TRAINS annotation schemes. With the help of domain specialists, we have constructed a hybrid multi-level annotation for a fragment of the Boston University Radio Speech Corpus which includes the following levels: segment, word, breath, ToBI, Tilt, Treebank, coreference and named entity. We show how annotation graphs can represent hybrid multi-level structures which derive from a diverse set of file formats. We also show how the approach facilitates substantive comparison of multiple annotations of a single signal based on different theoretical models. The discussion shows how annotation graphs open the door to wide-ranging integration of tools, formats and corpora.