Modern speech recognition systems rely on self-attention. Unfortunately, token mixing with self-attention takes quadratic time in the length of the speech utterance, slowing down inference as well as training and increasing memory consumption. Cheaper alternatives to self-attention for ASR have been developed, but fail to consistently reach the same level of accuracy. In practice, however, the self-attention weights of trained speech recognizers take the form of a global average over time. This paper, therefore, proposes a linear-time alternative to self-attention for speech recognition. It summarises a whole utterance with the mean over vectors for all time steps. This single summary is then combined with time-specific information. We call this method ``Summary Mixing''. Introducing Summary Mixing in state-of-the-art ASR models makes it feasible to preserve or exceed previous speech recognition performance while lowering the training and inference times by up to 27% and reducing the memory budget by a factor of two.
In recent research, slight performance improvement is observed from automatic speech recognition systems to audio-visual speech recognition systems in the end-to-end framework with low-quality videos. Unmatching convergence rates and specialized input representations between audio and visual modalities are considered to cause the problem. In this paper, we propose two novel techniques to improve audio-visual speech recognition (AVSR) under a pre-training and fine-tuning training framework. First, we explore the correlation between lip shapes and syllable-level subword units in Mandarin to establish good frame-level syllable boundaries from lip shapes. This enables accurate alignment of video and audio streams during visual model pre-training and cross-modal fusion. Next, we propose an audio-guided cross-modal fusion encoder (CMFE) neural network to utilize main training parameters for multiple cross-modal attention layers to make full use of modality complementarity. Experiments on the MISP2021-AVSR data set show the effectiveness of the two proposed techniques. Together, using only a relatively small amount of training data, the final system achieves better performances than state-of-the-art systems with more complex front-ends and back-ends.
In this paper, we aim to create weak alignment supervision from an existing hybrid system to aid the end-to-end modeling of automatic speech recognition. Towards this end, we use the existing hybrid ASR system to produce triphone alignments of the training audios. We then create a cross-entropy loss at a certain layer of the encoder using the derived alignments. In contrast to the general one-hot cross-entropy losses, here we use a cross-entropy loss with a label smoothing parameter to regularize the supervision. As a comparison, we also conduct the experiments with one-hot cross-entropy losses and CTC losses with loss weighting. The results show that placing the weak alignment supervision with the label smoothing parameter of 0.5 at the third encoder layer outperforms the other two approaches and leads to about 5\% relative WER reduction on the TED-LIUM 2 dataset over the baseline. We see similar improvements when applying the method out-of-the-box on a Tagalog end-to-end ASR system.
Speech-to-text translation pertains to the task of converting speech signals in a language to text in another language. It finds its application in various domains, such as hands-free communication, dictation, video lecture transcription, and translation, to name a few. Automatic Speech Recognition (ASR), as well as Machine Translation(MT) models, play crucial roles in traditional ST translation, enabling the conversion of spoken language in its original form to written text and facilitating seamless cross-lingual communication. ASR recognizes spoken words, while MT translates the transcribed text into the target language. Such disintegrated models suffer from cascaded error propagation and high resource and training costs. As a result, researchers have been exploring end-to-end (E2E) models for ST translation. However, to our knowledge, there is no comprehensive review of existing works on E2E ST. The present survey, therefore, discusses the work in this direction. Our attempt has been to provide a comprehensive review of models employed, metrics, and datasets used for ST tasks, providing challenges and future research direction with new insights. We believe this review will be helpful to researchers working on various applications of ST models.
Emotional well-being significantly influences mental health and overall quality of life. As therapy chatbots become increasingly prevalent, their ability to comprehend and respond empathetically to users' emotions remains limited. This paper addresses this limitation by proposing an approach to enhance therapy chatbots with auditory perception, enabling them to understand users' feelings and provide human-like empathy. The proposed method incorporates speech emotion recognition (SER) techniques using Convolutional Neural Network (CNN) models and the ShEMO dataset to accurately detect and classify negative emotions, including anger, fear, and sadness. The SER model achieves a validation accuracy of 88%, demonstrating its effectiveness in recognizing emotional states from speech signals. Furthermore, a recommender system is developed, leveraging the SER model's output to generate personalized recommendations for managing negative emotions, for which a new bilingual dataset was generated as well since there is no such dataset available for this task. The recommender model achieves an accuracy of 98% by employing a combination of global vectors for word representation (GloVe) and LSTM models. To provide a more immersive and empathetic user experience, a text-to-speech model called GlowTTS is integrated, enabling the therapy chatbot to audibly communicate the generated recommendations to users in both English and Persian. The proposed approach offers promising potential to enhance therapy chatbots by providing them with the ability to recognize and respond to users' emotions, ultimately improving the delivery of mental health support for both English and Persian-speaking users.
Advancements in deep neural networks have allowed automatic speech recognition (ASR) systems to attain human parity on several publicly available clean speech datasets. However, even state-of-the-art ASR systems experience performance degradation when confronted with adverse conditions, as a well-trained acoustic model is sensitive to variations in the speech domain, e.g., background noise. Intuitively, humans address this issue by relying on their linguistic knowledge: the meaning of ambiguous spoken terms is usually inferred from contextual cues thereby reducing the dependency on the auditory system. Inspired by this observation, we introduce the first open-source benchmark to utilize external large language models (LLMs) for ASR error correction, where N-best decoding hypotheses provide informative elements for true transcription prediction. This approach is a paradigm shift from the traditional language model rescoring strategy that can only select one candidate hypothesis as the output transcription. The proposed benchmark contains a novel dataset, HyPoradise (HP), encompassing more than 334,000 pairs of N-best hypotheses and corresponding accurate transcriptions across prevalent speech domains. Given this dataset, we examine three types of error correction techniques based on LLMs with varying amounts of labeled hypotheses-transcription pairs, which gains a significant word error rate (WER) reduction. Experimental evidence demonstrates the proposed technique achieves a breakthrough by surpassing the upper bound of traditional re-ranking based methods. More surprisingly, LLM with reasonable prompt and its generative capability can even correct those tokens that are missing in N-best list. We make our results publicly accessible for reproducible pipelines with released pre-trained models, thus providing a new evaluation paradigm for ASR error correction with LLMs.
In this work, we study the features extracted by English self-supervised learning (SSL) models in cross-lingual contexts and propose a new metric to predict the quality of feature representations. Using automatic speech recognition (ASR) as a downstream task, we analyze the effect of model size, training objectives, and model architecture on the models' performance as a feature extractor for a set of topologically diverse corpora. We develop a novel metric, the Phonetic-Syntax Ratio (PSR), to measure the phonetic and synthetic information in the extracted representations using deep generalized canonical correlation analysis. Results show the contrastive loss in the wav2vec2.0 objective facilitates more effective cross-lingual feature extraction. There is a positive correlation between PSR scores and ASR performance, suggesting that phonetic information extracted by monolingual SSL models can be used for downstream tasks in cross-lingual settings. The proposed metric is an effective indicator of the quality of the representations and can be useful for model selection.
When a speaker verification (SV) system operates far from the sound sourced, significant challenges arise due to the interference of noise and reverberation. Studies have shown that incorporating phonetic information into speaker embedding can improve the performance of text-independent SV. Inspired by this observation, we propose a joint-training speech recognition and speaker recognition (JTSS) framework to exploit phonetic content for far-field SV. The framework encourages speaker embeddings to preserve phonetic information by matching the frame-based feature maps of a speaker embedding network with wav2vec's vectors. The intuition is that phonetic information can preserve low-level acoustic dynamics with speaker information and thus partly compensate for the degradation due to noise and reverberation. Results show that the proposed framework outperforms the standard speaker embedding on the VOiCES Challenge 2019 evaluation set and the VoxCeleb1 test set. This indicates that leveraging phonetic information under far-field conditions is effective for learning robust speaker representations.
Joint rich and normalized automatic speech recognition (ASR), that produces transcriptions both with and without punctuation and capitalization, remains a challenge. End-to-end (E2E) ASR models offer both convenience and the ability to perform such joint transcription of speech. Training such models requires paired speech and rich text data, which is not widely available. In this paper, we compare two different approaches to train a stateless Transducer-based E2E joint rich and normalized ASR system, ready for streaming applications, with a limited amount of rich labeled data. The first approach uses a language model to generate pseudo-rich transcriptions of normalized training data. The second approach uses a single decoder conditioned on the type of the output. The first approach leads to E2E rich ASR which perform better on out-of-domain data, with up to 9% relative reduction in errors. The second approach demonstrates the feasibility of an E2E joint rich and normalized ASR system using as low as 5% rich training data with moderate (2.42% absolute) increase in errors.
Visual Speech Recognition (VSR) is the task of predicting spoken words from silent lip movements. VSR is regarded as a challenging task because of the insufficient information on lip movements. In this paper, we propose an Audio Knowledge empowered Visual Speech Recognition framework (AKVSR) to complement the insufficient speech information of visual modality by using audio modality. Different from the previous methods, the proposed AKVSR 1) utilizes rich audio knowledge encoded by a large-scale pretrained audio model, 2) saves the linguistic information of audio knowledge in compact audio memory by discarding the non-linguistic information from the audio through quantization, and 3) includes Audio Bridging Module which can find the best-matched audio features from the compact audio memory, which makes our training possible without audio inputs, once after the compact audio memory is composed. We validate the effectiveness of the proposed method through extensive experiments, and achieve new state-of-the-art performances on the widely-used datasets, LRS2 and LRS3.