Sequence-to-sequence models have achieved impressive results on various tasks.

Neural transducers (NT) provide an effective framework for speech streaming, demonstrating strong performance in automatic speech recognition (ASR). However, the application of NT to speech translation (ST) remains challenging, as existing approaches struggle with word reordering and performance degradation when jointly modeling ASR and ST, resulting in a gap with attention-based encoder-decoder (AED) models. Existing NT-based ST approaches also suffer from high computational training costs. To address these issues, we propose HENT-SRT (Hierarchical Efficient Neural Transducer for Speech Recognition and Translation), a novel framework that factorizes ASR and translation tasks to better handle reordering. To ensure robust ST while preserving ASR performance, we use self-distillation with CTC consistency regularization. Moreover, we improve computational efficiency by incorporating best practices from ASR transducers, including a down-sampled hierarchical encoder, a stateless predictor, and a pruned transducer loss to reduce training complexity. Finally, we introduce a blank penalty during decoding, reducing deletions and improving translation quality. Our approach is evaluated on three conversational datasets Arabic, Spanish, and Mandarin achieving new state-of-the-art performance among NT models and substantially narrowing the gap with AED-based systems.

While large language models (LLMs) have been applied to automatic speech recognition (ASR), the task of making the model streamable remains a challenge. This paper proposes a novel model architecture, Transducer-Llama, that integrates LLMs into a Factorized Transducer (FT) model, naturally enabling streaming capabilities. Furthermore, given that the large vocabulary of LLMs can cause data sparsity issue and increased training costs for spoken language systems, this paper introduces an efficient vocabulary adaptation technique to align LLMs with speech system vocabularies. The results show that directly optimizing the FT model with a strong pre-trained LLM-based predictor using the RNN-T loss yields some but limited improvements over a smaller pre-trained LM predictor. Therefore, this paper proposes a weak-to-strong LM swap strategy, using a weak LM predictor during RNN-T loss training and then replacing it with a strong LLM. After LM replacement, the minimum word error rate (MWER) loss is employed to finetune the integration of the LLM predictor with the Transducer-Llama model. Experiments on the LibriSpeech and large-scale multi-lingual LibriSpeech corpora show that the proposed streaming Transducer-Llama approach gave a 17% relative WER reduction (WERR) over a strong FT baseline and a 32% WERR over an RNN-T baseline.

Self-supervised learning, such as with the wav2vec 2.0 framework significantly improves the accuracy of end-to-end automatic speech recognition (ASR). Wav2vec 2.0 has been applied to single-channel end-to-end ASR models. In this work, we explored a self-supervised learning method for a multi-channel end-to-end ASR model based on the wav2vec 2.0 framework. As the multi-channel end-to-end ASR model, we focused on a multi-channel neural transducer. In pre-training, we compared three different methods for feature quantization to train a multi-channel conformer audio encoder: joint quantization, feature-wise quantization and channel-wise quantization. In fine-tuning, we trained the multi-channel conformer-transducer. All experiments were conducted using the far-field in-house and CHiME-4 datasets. The results of the experiments showed that feature-wise quantization was the most effective among the methods. We observed a 66% relative reduction in character error rate compared with the model without any pre-training for the far-field in-house dataset.

Language-agnostic many-to-one end-to-end speech translation models can convert audio signals from different source languages into text in a target language. These models do not need source language identification, which improves user experience. In some cases, the input language can be given or estimated. Our goal is to use this additional language information while preserving the quality of the other languages. We accomplish this by introducing a simple and effective linear input network. The linear input network is initialized as an identity matrix, which ensures that the model can perform as well as, or better than, the original model. Experimental results show that the proposed method can successfully enhance the specified language, while keeping the language-agnostic ability of the many-to-one ST models.

While the neural transducer is popular for online speech recognition, simultaneous speech translation (SST) requires both streaming and re-ordering capabilities. This paper presents the LS-Transducer-SST, a label-synchronous neural transducer for SST, which naturally possesses these two properties. The LS-Transducer-SST dynamically decides when to emit translation tokens based on an Auto-regressive Integrate-and-Fire (AIF) mechanism. A latency-controllable AIF is also proposed, which can control the quality-latency trade-off either only during decoding, or it can be used in both decoding and training. The LS-Transducer-SST can naturally utilise monolingual text-only data via its prediction network which helps alleviate the key issue of data sparsity for E2E SST. During decoding, a chunk-based incremental joint decoding technique is designed to refine and expand the search space. Experiments on the Fisher-CallHome Spanish (Es-En) and MuST-C En-De data show that the LS-Transducer-SST gives a better quality-latency trade-off than existing popular methods. For example, the LS-Transducer-SST gives a 3.1/2.9 point BLEU increase (Es-En/En-De) relative to CAAT at a similar latency and a 1.4 s reduction in average lagging latency with similar BLEU scores relative to Wait-k.

Sequence-to-sequence models have achieved impressive results on various tasks. However, they are unsuitable for tasks that require incremental predictions to be made as more data arrives or tasks that have long input sequences and output sequences. This is because they generate an output sequence conditioned on an entire input sequence. In this paper, we present a Neural Transducer that can make incremental predictions as more input arrives, without redoing the entire computation. Unlike sequence-to-sequence models, the Neural Transducer computes the next-step distribution conditioned on the partially observed input sequence and the partially generated sequence.

We propose a novel text-to-speech (TTS) framework centered around a neural transducer. Our approach divides the whole TTS pipeline into semantic-level sequence-to-sequence (seq2seq) modeling and fine-grained acoustic modeling stages, utilizing discrete semantic tokens obtained from wav2vec2.0 embeddings. For a robust and efficient alignment modeling, we employ a neural transducer named token transducer for the semantic token prediction, benefiting from its hard monotonic alignment constraints. Subsequently, a non-autoregressive (NAR) speech generator efficiently synthesizes waveforms from these semantic tokens. Additionally, a reference speech controls temporal dynamics and acoustic conditions at each stage. This decoupled framework reduces the training complexity of TTS while allowing each stage to focus on semantic and acoustic modeling. Our experimental results on zero-shot adaptive TTS demonstrate that our model surpasses the baseline in terms of speech quality and speaker similarity, both objectively and subjectively. We also delve into the inference speed and prosody control capabilities of our approach, highlighting the potential of neural transducers in TTS frameworks.

We introduce a text-to-speech(TTS) framework based on a neural transducer. We use discretized semantic tokens acquired from wav2vec2.0 embeddings, which makes it easy to adopt a neural transducer for the TTS framework enjoying its monotonic alignment constraints. The proposed model first generates aligned semantic tokens using the neural transducer, then synthesizes a speech sample from the semantic tokens using a non-autoregressive(NAR) speech generator. This decoupled framework alleviates the training complexity of TTS and allows each stage to focus on 1) linguistic and alignment modeling and 2) fine-grained acoustic modeling, respectively. Experimental results on the zero-shot adaptive TTS show that the proposed model exceeds the baselines in speech quality and speaker similarity via objective and subjective measures. We also investigate the inference speed and prosody controllability of our proposed model, showing the potential of the neural transducer for TTS frameworks.

In this work, we investigate the effect of language models (LMs) with different context lengths and label units (phoneme vs. word) used in sequence discriminative training for phoneme-based neural transducers. Both lattice-free and N-best-list approaches are examined. For lattice-free methods with phoneme-level LMs, we propose a method to approximate the context history to employ LMs with full-context dependency. This approximation can be extended to arbitrary context length and enables the usage of word-level LMs in lattice-free methods. Moreover, a systematic comparison is conducted across lattice-free and N-best-list-based methods. Experimental results on Librispeech show that using the word-level LM in training outperforms the phoneme-level LM. Besides, we find that the context size of the LM used for probability computation has a limited effect on performance. Moreover, our results reveal the pivotal importance of the hypothesis space quality in sequence discriminative training.