Multilingual self-supervised speech representations improve the speech recognition of low-resource African languages with codeswitching

While many speakers of low-resource languages regularly code-switch between their languages and other regional languages or English, datasets of codeswitched speech are too small to train bespoke acoustic models from scratch or do language model rescoring. Here we propose finetuning self-supervised speech representations such as wav2vec 2.0 XLSR to recognize code-switched data. We find that finetuning self-supervised multilingual representations and augmenting them with n-gram language models trained from transcripts reduces absolute word error rates by up to 20% compared to baselines of hybrid models trained from scratch on code-switched data. Our findings suggest that in circumstances with limited training data finetuning self-supervised representations is a better performing and viable solution.

Adapting Large Language Model with Speech for Fully Formatted End-to-End Speech Recognition

Most end-to-end (E2E) speech recognition models are composed of encoder and decoder blocks that perform acoustic and language modeling functions. Pretrained large language models (LLMs) have the potential to improve the performance of E2E ASR. However, integrating a pretrained language model into an E2E speech recognition model has shown limited benefits due to the mismatches between text-based LLMs and those used in E2E ASR. In this paper, we explore an alternative approach by adapting a pretrained LLMs to speech. Our experiments on fully-formatted E2E ASR transcription tasks across various domains demonstrate that our approach can effectively leverage the strengths of pretrained LLMs to produce more readable ASR transcriptions. Our model, which is based on the pretrained large language models with either an encoder-decoder or decoder-only structure, surpasses strong ASR models such as Whisper, in terms of recognition error rate, considering formats like punctuation and capitalization as well.

Bring the Noise: Introducing Noise Robustness to Pretrained Automatic Speech Recognition

In recent research, in the domain of speech processing, large End-to-End (E2E) systems for Automatic Speech Recognition (ASR) have reported state-of-the-art performance on various benchmarks. These systems intrinsically learn how to handle and remove noise conditions from speech. Previous research has shown, that it is possible to extract the denoising capabilities of these models into a preprocessor network, which can be used as a frontend for downstream ASR models. However, the proposed methods were limited to specific fully convolutional architectures. In this work, we propose a novel method to extract the denoising capabilities, that can be applied to any encoder-decoder architecture. We propose the Cleancoder preprocessor architecture that extracts hidden activations from the Conformer ASR model and feeds them to a decoder to predict denoised spectrograms. We train our pre-processor on the Noisy Speech Database (NSD) to reconstruct denoised spectrograms from noisy inputs. Then, we evaluate our model as a frontend to a pretrained Conformer ASR model as well as a frontend to train smaller Conformer ASR models from scratch. We show that the Cleancoder is able to filter noise from speech and that it improves the total Word Error Rate (WER) of the downstream model in noisy conditions for both applications.

Whisper-MCE: Whisper Model Finetuned for Better Performance with Mixed Languages

Recently Whisper has approached human-level robustness and accuracy in English automatic speech recognition (ASR), while in minor language and mixed language speech recognition, there remains a compelling need for further improvement. In this work, we present the impressive results of Whisper-MCE, our finetuned Whisper model, which was trained using our self-collected dataset, Mixed Cantonese and English audio dataset (MCE). Meanwhile, considering word error rate (WER) poses challenges when it comes to evaluating its effectiveness in minor language and mixed-language contexts, we present a novel rating mechanism. By comparing our model to the baseline whisper-large-v2 model, we demonstrate its superior ability to accurately capture the content of the original audio, achieve higher recognition accuracy, and exhibit faster recognition speed. Notably, our model outperforms other existing models in the specific task of recognizing mixed language.

Soft Random Sampling: A Theoretical and Empirical Analysis

Soft random sampling (SRS) is a simple yet effective approach for efficient training of large-scale deep neural networks when dealing with massive data. SRS selects a subset uniformly at random with replacement from the full data set in each epoch. In this paper, we conduct a theoretical and empirical analysis of SRS. First, we analyze its sampling dynamics including data coverage and occupancy. Next, we investigate its convergence with non-convex objective functions and give the convergence rate. Finally, we provide its generalization performance. We empirically evaluate SRS for image recognition on CIFAR10 and automatic speech recognition on Librispeech and an in-house payload dataset to demonstrate its effectiveness. Compared to existing coreset-based data selection methods, SRS offers a better accuracy-efficiency trade-off. Especially on real-world industrial scale data sets, it is shown to be a powerful training strategy with significant speedup and competitive performance with almost no additional computing cost.

Federated Representation Learning for Automatic Speech Recognition

Federated Learning (FL) is a privacy-preserving paradigm, allowing edge devices to learn collaboratively without sharing data. Edge devices like Alexa and Siri are prospective sources of unlabeled audio data that can be tapped to learn robust audio representations. In this work, we bring Self-supervised Learning (SSL) and FL together to learn representations for Automatic Speech Recognition respecting data privacy constraints. We use the speaker and chapter information in the unlabeled speech dataset, Libri-Light, to simulate non-IID speaker-siloed data distributions and pre-train an LSTM encoder with the Contrastive Predictive Coding framework with FedSGD. We show that the pre-trained ASR encoder in FL performs as well as a centrally pre-trained model and produces an improvement of 12-15% (WER) compared to no pre-training. We further adapt the federated pre-trained models to a new language, French, and show a 20% (WER) improvement over no pre-training.

Language-Routing Mixture of Experts for Multilingual and Code-Switching Speech Recognition

Multilingual speech recognition for both monolingual and code-switching speech is a challenging task. Recently, based on the Mixture of Experts (MoE), many works have made good progress in multilingual and code-switching ASR, but present huge computational complexity with the increase of supported languages. In this work, we propose a computation-efficient network named Language-Routing Mixture of Experts (LR-MoE) for multilingual and code-switching ASR. LR-MoE extracts language-specific representations through the Mixture of Language Experts (MLE), which is guided to learn by a frame-wise language routing mechanism. The weight-shared frame-level language identification (LID) network is jointly trained as the shared pre-router of each MoE layer. Experiments show that the proposed method significantly improves multilingual and code-switching speech recognition performances over baseline with comparable computational efficiency.

Text-Only Domain Adaptation for End-to-End Speech Recognition through Down-Sampling Acoustic Representation

Mapping two modalities, speech and text, into a shared representation space, is a research topic of using text-only data to improve end-to-end automatic speech recognition (ASR) performance in new domains. However, the length of speech representation and text representation is inconsistent. Although the previous method up-samples the text representation to align with acoustic modality, it may not match the expected actual duration. In this paper, we proposed novel representations match strategy through down-sampling acoustic representation to align with text modality. By introducing a continuous integrate-and-fire (CIF) module generating acoustic representations consistent with token length, our ASR model can learn unified representations from both modalities better, allowing for domain adaptation using text-only data of the target domain. Experiment results of new domain data demonstrate the effectiveness of the proposed method.

Boosting Norwegian Automatic Speech Recognition

In this paper, we present several baselines for automatic speech recognition (ASR) models for the two official written languages in Norway: Bokm{\aa}l and Nynorsk. We compare the performance of models of varying sizes and pre-training approaches on multiple Norwegian speech datasets. Additionally, we measure the performance of these models against previous state-of-the-art ASR models, as well as on out-of-domain datasets. We improve the state of the art on the Norwegian Parliamentary Speech Corpus (NPSC) from a word error rate (WER) of 17.10\% to 7.60\%, with models achieving 5.81\% for Bokm{\aa}l and 11.54\% for Nynorsk. We also discuss the challenges and potential solutions for further improving ASR models for Norwegian.

Harnessing the Zero-Shot Power of Instruction-Tuned Large Language Model in End-to-End Speech Recognition

We present a novel integration of an instruction-tuned large language model (LLM) and end-to-end automatic speech recognition (ASR). Modern LLMs can perform a wide range of linguistic tasks within zero-shot learning when provided with a precise instruction or a prompt to guide the text generation process towards the desired task. We explore using this zero-shot capability of LLMs to extract linguistic information that can contribute to improving ASR performance. Specifically, we direct an LLM to correct grammatical errors in an ASR hypothesis and harness the embedded linguistic knowledge to conduct end-to-end ASR. The proposed model is built on the hybrid connectionist temporal classification (CTC) and attention architecture, where an instruction-tuned LLM (i.e., Llama2) is employed as a front-end of the decoder. An ASR hypothesis, subject to correction, is obtained from the encoder via CTC decoding, which is then fed into the LLM along with an instruction. The decoder subsequently takes as input the LLM embeddings to perform sequence generation, incorporating acoustic information from the encoder output. Experimental results and analyses demonstrate that the proposed integration yields promising performance improvements, and our approach largely benefits from LLM-based rescoring.