Connectionist Temporal Classification (CTC) models are popular for their balance between speed and performance for Automatic Speech Recognition (ASR). However, these CTC models still struggle in other areas, such as personalization towards custom words. A recent approach explores Contextual Adapters, wherein an attention-based biasing model for CTC is used to improve the recognition of custom entities. While this approach works well with enough data, we showcase that it isn't an effective strategy for low-resource languages. In this work, we propose a supervision loss for smoother training of the Contextual Adapters. Further, we explore a multilingual strategy to improve performance with limited training data. Our method achieves 48% F1 improvement in retrieving unseen custom entities for a low-resource language. Interestingly, as a by-product of training the Contextual Adapters, we see a 5-11% Word Error Rate (WER) reduction in the performance of the base CTC model as well.
Swarm Intelligence algorithms have gained significant attention in recent years as a means of solving complex and non-deterministic problems. These algorithms are inspired by the collective behavior of natural creatures, and they simulate this behavior to develop intelligent agents for computational tasks. One such algorithm is Ant Colony Optimization (ACO), which is inspired by the foraging behavior of ants and their pheromone laying mechanism. ACO is used for solving difficult problems that are discrete and combinatorial in nature. Part-of-Speech (POS) tagging is a fundamental task in natural language processing that aims to assign a part-of-speech role to each word in a sentence. In this research paper, proposed a high-performance POS-tagging method based on ACO called ACO-tagger. This method achieved a high accuracy rate of 96.867%, outperforming several state-of-the-art methods. The proposed method is fast and efficient, making it a viable option for practical applications.
We introduce EM-Network, a novel self-distillation approach that effectively leverages target information for supervised sequence-to-sequence (seq2seq) learning. In contrast to conventional methods, it is trained with oracle guidance, which is derived from the target sequence. Since the oracle guidance compactly represents the target-side context that can assist the sequence model in solving the task, the EM-Network achieves a better prediction compared to using only the source input. To allow the sequence model to inherit the promising capability of the EM-Network, we propose a new self-distillation strategy, where the original sequence model can benefit from the knowledge of the EM-Network in a one-stage manner. We conduct comprehensive experiments on two types of seq2seq models: connectionist temporal classification (CTC) for speech recognition and attention-based encoder-decoder (AED) for machine translation. Experimental results demonstrate that the EM-Network significantly advances the current state-of-the-art approaches, improving over the best prior work on speech recognition and establishing state-of-the-art performance on WMT'14 and IWSLT'14.
Pre-trained speech encoders have been central to pushing state-of-the-art results across various speech understanding and generation tasks. Nonetheless, the capabilities of these encoders in low-resource settings are yet to be thoroughly explored. To address this, we conduct a comprehensive set of experiments using a representative set of 3 state-of-the-art encoders (Wav2vec2, WavLM, Whisper) in the low-resource setting across 7 speech understanding and generation tasks. We provide various quantitative and qualitative analyses on task performance, convergence speed, and representational properties of the encoders. We observe a connection between the pre-training protocols of these encoders and the way in which they capture information in their internal layers. In particular, we observe the Whisper encoder exhibits the greatest low-resource capabilities on content-driven tasks in terms of performance and convergence speed.
Visual speech recognition models extract visual features in a hierarchical manner. At the lower level, there is a visual front-end with a limited temporal receptive field that processes the raw pixels depicting the lips or faces. At the higher level, there is an encoder that attends to the embeddings produced by the front-end over a large temporal receptive field. Previous work has focused on improving the visual front-end of the model to extract more useful features for speech recognition. Surprisingly, our work shows that complex visual front-ends are not necessary. Instead of allocating resources to a sophisticated visual front-end, we find that a linear visual front-end paired with a larger Conformer encoder results in lower latency, more efficient memory usage, and improved WER performance. We achieve a new state-of-the-art of $12.8\%$ WER for visual speech recognition on the TED LRS3 dataset, which rivals the performance of audio-only models from just four years ago.
Language models have been successfully used to model natural signals, such as images, speech, and music. A key component of these models is a high quality neural compression model that can compress high-dimensional natural signals into lower dimensional discrete tokens. To that end, we introduce a high-fidelity universal neural audio compression algorithm that achieves ~90x compression of 44.1 KHz audio into tokens at just 8kbps bandwidth. We achieve this by combining advances in high-fidelity audio generation with better vector quantization techniques from the image domain, along with improved adversarial and reconstruction losses. We compress all domains (speech, environment, music, etc.) with a single universal model, making it widely applicable to generative modeling of all audio. We compare with competing audio compression algorithms, and find our method outperforms them significantly. We provide thorough ablations for every design choice, as well as open-source code and trained model weights. We hope our work can lay the foundation for the next generation of high-fidelity audio modeling.
This paper proposes a self-regularised minimum latency training (SR-MLT) method for streaming Transformer-based automatic speech recognition (ASR) systems. In previous works, latency was optimised by truncating the online attention weights based on the hard alignments obtained from conventional ASR models, without taking into account the potential loss of ASR accuracy. On the contrary, here we present a strategy to obtain the alignments as a part of the model training without external supervision. The alignments produced by the proposed method are dynamically regularised on the training data, such that the latency reduction does not result in the loss of ASR accuracy. SR-MLT is applied as a fine-tuning step on the pre-trained Transformer models that are based on either monotonic chunkwise attention (MoChA) or cumulative attention (CA) algorithms for online decoding. ASR experiments on the AIShell-1 and Librispeech datasets show that when applied on a decent pre-trained MoChA or CA baseline model, SR-MLT can effectively reduce the latency with the relative gains ranging from 11.8% to 39.5%. Furthermore, we also demonstrate that under certain accuracy levels, the models trained with SR-MLT can achieve lower latency when compared to those supervised using external hard alignments.
Existing self-supervised pre-trained speech models have offered an effective way to leverage massive unannotated corpora to build good automatic speech recognition (ASR). However, many current models are trained on a clean corpus from a single source, which tends to do poorly when noise is present during testing. Nonetheless, it is crucial to overcome the adverse influence of noise for real-world applications. In this work, we propose a novel training framework, called deHuBERT, for noise reduction encoding inspired by H. Barlow's redundancy-reduction principle. The new framework improves the HuBERT training algorithm by introducing auxiliary losses that drive the self- and cross-correlation matrix between pairwise noise-distorted embeddings towards identity matrix. This encourages the model to produce noise-agnostic speech representations. With this method, we report improved robustness in noisy environments, including unseen noises, without impairing the performance on the clean set.
Data augmentations are known to improve robustness in speech-processing tasks. In this study, we summarize and compare different data augmentation strategies using S3PRL toolkit. We explore how HuBERT and wav2vec perform using different augmentation techniques (SpecAugment, Gaussian Noise, Speed Perturbation) for Phoneme Recognition (PR) and Automatic Speech Recognition (ASR) tasks. We evaluate model performance in terms of phoneme error rate (PER) and word error rate (WER). From the experiments, we observed that SpecAugment slightly improves the performance of HuBERT and wav2vec on the original dataset. Also, we show that models trained using the Gaussian Noise and Speed Perturbation dataset are more robust when tested with augmented test sets.
While Transformer has become the de-facto standard for speech, modeling upon the fine-grained frame-level features remains an open challenge of capturing long-distance dependencies and distributing the attention weights. We propose \textit{Progressive Down-Sampling} (PDS) which gradually compresses the acoustic features into coarser-grained units containing more complete semantic information, like text-level representation. In addition, we develop a representation fusion method to alleviate information loss that occurs inevitably during high compression. In this way, we compress the acoustic features into 1/32 of the initial length while achieving better or comparable performances on the speech recognition task. And as a bonus, it yields inference speedups ranging from 1.20$\times$ to 1.47$\times$. By reducing the modeling burden, we also achieve competitive results when training on the more challenging speech translation task.