Siqi Ouyang, Shuoyang Ding, Oleksii Hrinchuk, Vitaly Lavrukhin, Brian Yan, Boris Ginsburg, Lei Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Mehrzad Samadi, Aleksander Ficek, Sean Narenthiran, Siddhartha Jain, Wasi Uddin Ahmad, Somshubra Majumdar, Vahid Noroozi, Boris Ginsburg. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Zhen Wan, Chao-Han Huck Yang, Jinchuan Tian, Hanrong Ye, Ankita Pasad, Szu-Wei Fu, Arushi Goel, Ryo Hachiuma, Shizhe Diao, Kunal Dhawan, Sreyan Ghosh, Yusuke Hirota, Zhehuai Chen, Rafael Valle, Chenhui Chu, Shinji Watanabe, Boris Ginsburg, Yu-Chiang Frank Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recognizing specific key phrases is an essential task for contextualized Automatic Speech Recognition (ASR). However, most existing context-biasing approaches have limitations associated with the necessity of additional model training, significantly slow down the decoding process, or constrain the choice of the ASR system type. This paper proposes a universal ASR context-biasing framework that supports all major types: CTC, Transducers, and Attention Encoder-Decoder models. The framework is based on a GPU-accelerated word boosting tree, which enables it to be used in shallow fusion mode for greedy and beam search decoding without noticeable speed degradation, even with a vast number of key phrases (up to 20 K items). The obtained results showed high efficiency of the proposed method, surpassing the considered open-source context-biasing approaches in accuracy and decoding speed. Our context-biasing framework is open-sourced as a part of the NeMo toolkit.

We present the system demonstration and opensource code release of a novel, data-efficient framework that converts any standard text Large Language Model (LLM) into a full-duplex end-to-end (E2E) speech-to-speech (S2S) model, for building conversational voice agents. Our new modeling method enables any LLMs to simultaneously listen and speak without requiring extensive speech-text pretraining. Moreover, we demonstrate how to put together a low-latency and full-duplex voice agent with open-source modeling, inference optimization, and serving solutions. This work significantly lowers the barrier to entry for developing low-latency, human-like voice agents by providing a generalizable, end-to-end solution built on open-source technologies.

While beam search improves speech recognition quality over greedy decoding, standard implementations are slow, often sequential, and CPU-bound. To fully leverage modern hardware capabilities, we present a novel open-source FlexCTC toolkit for fully GPU-based beam decoding, designed for Connectionist Temporal Classification (CTC) models. Developed entirely in Python and PyTorch, it offers a fast, user-friendly, and extensible alternative to traditional C++, CUDA, or WFST-based decoders. The toolkit features a high-performance, fully batched GPU implementation with eliminated CPU-GPU synchronization and minimized kernel launch overhead via CUDA Graphs. It also supports advanced contextualization techniques, including GPU-powered N-gram language model fusion and phrase-level boosting. These features enable accurate and efficient decoding, making them suitable for both research and production use.

Attention encoder-decoder architecture is the backbone of several top performing foundation speech models: Whisper, Seamless, OWSM, and Canary-1B. However, reported compute requirements are prohibitive for many researchers. In this work, we seek to improve both training and inference efficiency. We argue that a major detrimental factor is the sampling strategy of sequential data. Negligence in mini-batch sampling leads to over 50% computation spent on padding. Using improved 2D bucketing combined with a batch size optimizer, we achieve 5x increase in average batch sizes for Canary-1B training, allowing 4x less GPUs or 2x shorter training time. Finally, the major inference bottleneck lies in autoregressive decoder steps. We show that transferring parameters from decoder to encoder results in 3x inference speedup while preserving accuracy. The training code and models are open-source with permissive licenses.

Krishna C Puvvada, Faisal Ladhak, Santiago Akle Serano, Cheng-Ping Hsieh, Shantanu Acharya, Somshubra Majumdar, Fei Jia, Samuel Kriman, Simeng Sun, Dima Rekesh, Boris Ginsburg. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Kuang-Da Wang, Shuoyang Ding, Chao-Han Huck Yang, Ping-Chun Hsieh, Wen-Chih Peng, Vitaly Lavrukhin, Boris Ginsburg. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Despite Arabic being one of the most widely spoken languages, the development of Arabic Automatic Speech Recognition (ASR) systems faces significant challenges due to the language’s complexity, and only a limited number of public Arabic ASR models exist. While much of the focus has been on Modern Standard Arabic (MSA), there is considerably less attention given to the variations within the language. This paper introduces a universal methodology for Arabic speech and text processing designed to address unique challenges of the language. Using this methodology, we train two novel models based on the FastConformer architecture: one designed specifically for MSA and the other, the first unified public model for both MSA and Classical Arabic (CA). The MSA model sets a new benchmark with state-of-the-art (SOTA) performance on related datasets, while the unified model achieves SOTA accuracy with diacritics for CA while maintaining strong performance for MSA. To promote reproducibility, we open-source the models and their training recipes.

Large language models (LLMs) have demonstrated remarkable advancements in language understanding and generation. Building on the success of text-based LLMs, recent research has adapted these models to use speech embeddings for prompting, resulting in Speech-LLM models that exhibit strong performance in automatic speech recognition (ASR) and automatic speech translation (AST). In this work, we propose a novel approach to leverage ASR transcripts as prompts for AST in a Speech-LLM built on an encoder-decoder text LLM. The Speech-LLM model consists of a speech encoder and an encoder-decoder structure Megatron-T5. By first decoding speech to generate ASR transcripts and subsequently using these transcripts along with encoded speech for prompting, we guide the speech translation in a two-step process like chain-of-thought (CoT) prompting. Low-rank adaptation (LoRA) is used for the T5 LLM for model adaptation and shows superior performance to full model fine-tuning. Experimental results show that the proposed CoT prompting significantly improves AST performance, achieving an average increase of 2.4 BLEU points across 6 En→X or X→En AST tasks compared to speech prompting alone. Additionally, compared to a related CoT prediction method that predicts a concatenated sequence of ASR and AST transcripts, our method performs better by an average of 2 BLEU points.

Self-supervised learning (SSL) has been proved to benefit a wide range of speech processing tasks, such as speech recognition/translation, speaker verification and diarization, etc. However, most of current speech SSL approaches are computationally expensive. In this paper, we introduce a simplified and more efficient SSL framework, termed as NeMo Encoder for Speech Tasks (NEST). Specifically, we adopt the FastConformer architecture with 8x sub-sampling rate, which is faster than Transformer or Conformer architectures. Instead of clusteringbased quantization, we use fixed random projection for its simplicity and effectiveness. We also implement a generalized noisy speech augmentation that teaches the model to disentangle the main speaker from noise or other speakers. Experiments show that NEST improves over existing self-supervised models and achieves new state-of-the-art performance on a variety of speech processing tasks, such as speech recognition/translation, speaker diarization, spoken language understanding, etc. Code and checkpoints are publicly available via NVIDIA NeMo framework123.

Recent end-to-end speech language models (SLMs) have expanded upon the capabilities of large language models (LLMs) by incorporating pre-trained speech models. However, these SLMs often undergo extensive speech instruction-tuning to bridge the gap between speech and text modalities. This requires significant annotation efforts and risks catastrophic forgetting of the original language capabilities. In this work, we present a simple yet effective automatic process for creating speech-text pair data that carefully injects speech paralinguistic understanding abilities into SLMs while preserving the inherent language capabilities of the text-based LLM. Our model demonstrates general capabilities for speech-related tasks without the need for speech instruction-tuning data, achieving impressive performance on Dynamic-SUPERB and AIR-Bench-Chat benchmarks. Furthermore, our model exhibits the ability to follow complex instructions derived from LLMs, such as specific output formatting and chain-of-thought reasoning. Our approach not only enhances the versatility and effectiveness of SLMs but also reduces reliance on extensive annotated datasets, paving the way for more efficient and capable speech understanding systems.1

We propose a novel end-to-end multi-talker automatic speech recognition (ASR) framework that enables both multi-speaker (MS) ASR and target-speaker (TS) ASR. Our proposed model is trained in a fully end-to-end manner, incorporating speaker supervision from a pre-trained speaker diarization module. We introduce an intuitive yet effective method for masking ASR encoder activations using output from the speaker supervision module, a technique we term Meta-Cat (meta-information concatenation), that can be applied to both MS-ASR and TS-ASR. Our results demonstrate that the proposed architecture achieves competitive performance in both MS-ASR and TS-ASR tasks, without the need for traditional methods, such as neural mask estimation or masking at the audio or feature level. Furthermore, we demonstrate a glimpse of a unified dual-task model which can efficiently handle both MS-ASR and TS-ASR tasks. Thus, this work illustrates that a robust end-to-end multi-talker ASR framework can be implemented with a streamlined architecture, obviating the need for the complex speaker filtering mechanisms employed in previous studies.

A rising interest in the modality extension of foundation language models warrants discussion on the most effective, and efficient, multimodal training approach. This work focuses on neural machine translation (NMT) and proposes a joint multimodal training regime of Speech-LLM to include automatic speech translation (AST). We investigate two different foundation model architectures, decoder-only GPT and encoder-decoder T5, extended with Canary-1B’s speech encoder. To handle joint multimodal training, we propose a novel training framework called EMMeTT. EMMeTT improves training efficiency with the following: balanced sampling across languages, datasets, and modalities; efficient sequential data iteration; and a novel 2D bucketing scheme for multimodal data, complemented by a batch size optimizer (OOMptimizer). We show that a multimodal training consistently helps with both architectures. Moreover, SALM-T5 trained with EMMeTT retains the original NMT capability while outperforming AST baselines on four-language subsets of FLORES and FLEURS. The resultant Multimodal Translation Model produces strong text and speech translation results at the same time.

We propose a novel neural network architecture, the normalized Transformer (nGPT) with representation learning on the hypersphere. In nGPT, all vectors forming the embeddings, MLP, attention matrices and hidden states are unit norm normalized. The input stream of tokens travels on the surface of a hypersphere, with each layer contributing a displacement towards the target output predictions. These displacements are defined by the MLP and attention blocks, whose vector components also reside on the same hypersphere. Experiments show that nGPT learns much faster, reducing the number of training steps required to achieve the same accuracy by a factor of 4 to 20, depending on the sequence length.

We present Hybrid-Autoregressive INference TrANsducers (HAINAN), a novel architecture for speech recognition that extends the Token-and-Duration Transducer (TDT) model. Trained with randomly masked predictor network outputs, HAINAN supports both autoregressive inference with all network components and non-autoregressive inference without the predictor. Additionally, we propose a novel semi-autoregressive inference method that first generates an initial hypothesis using non-autoregressive inference, followed by refinement steps where each token prediction is regenerated using parallelized autoregression on the initial hypothesis. Experiments on multiple datasets across different languages demonstrate that HAINAN achieves efficiency parity with CTC in non-autoregressive mode and with TDT in autoregressive mode. In terms of accuracy, autoregressive HAINAN achieves parity with TDT and RNN-T, while non-autoregressive HAINAN significantly outperforms CTC. Semi-autoregressive inference further enhances the model's accuracy with minimal computational overhead, and even outperforms TDT results in some cases. These results highlight HAINAN's flexibility in balancing accuracy and speed, positioning it as a strong candidate for real-world speech recognition applications.

Inference with Transformer-based Large Language Models (LLMs) on long sequences is both costly and slow due to the quadratic complexity of the self-attention mechanism. We introduce Star Attention, a two-phase block-sparse approximation that improves computational efficiency by sharding attention across multiple hosts while minimizing communication overhead. In the first phase, the context is processed using blockwise-local attention across hosts, in parallel. In the second phase, query and response tokens attend to all prior cached tokens through sequence-global attention. Star Attention integrates seamlessly with most Transformer-based LLMs trained with global attention, reducing memory requirements and inference time by up to 11x while preserving 97-100% of accuracy.

Sortformer is an encoder-based speaker diarization model designed for supervising speaker tagging in speech-to-text models. Instead of relying solely on permutation invariant loss (PIL), Sortformer introduces Sort Loss to resolve the permutation problem, either independently or in tandem with PIL. In addition, we propose a streamlined multi-speaker speech-to-text architecture that leverages Sortformer for speaker supervision, embedding speaker labels into the encoder using sinusoidal kernel functions. This design addresses the speaker permutation problem through sorted objectives, effectively bridging timestamps and tokens to supervise speaker labels in the output transcriptions. Experiments demonstrate that Sort Loss can boost speaker diarization performance, and incorporating the speaker supervision from Sortformer improves multi-speaker transcription accuracy. We anticipate that the proposed Sortformer and multi-speaker architecture will enable the seamless integration of speaker tagging capabilities into foundational speech-to-text systems and multimodal large language models (LLMs), offering an easily adoptable and user-friendly mechanism to enhance their versatility and performance in speaker-aware tasks. The code and trained models are made publicly available through the NVIDIA NeMo Framework.

Siqi Ouyang, Oleksii Hrinchuk, Zhehuai Chen, Vitaly Lavrukhin, Jagadeesh Balam, Lei Li, Boris Ginsburg. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Yifan Peng, Krishna C Puvvada, Zhehuai Chen, Piotr Zelasko, He Huang, Kunal Dhawan, Ke Hu, Shinji Watanabe, Jagadeesh Balam, Boris Ginsburg. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Text normalization - the conversion of text from written to spoken form - is traditionally assumed to be an ill-formed task for language modeling. In this work, we argue otherwise. We empirically show the capacity of Large-Language Models (LLM) for text normalization in few-shot scenarios. Combining self-consistency reasoning with linguistic-informed prompt engineering, we find LLM-based text normalization to achieve error rates approximately 40% lower than production-level normalization systems. Further, upon error analysis, we note key limitations in the conventional design of text normalization tasks. We create a new taxonomy of text normalization errors and apply it to results from GPT-3.5-Turbo and GPT-4.0. Through this new framework, we identify strengths and weaknesses of LLM-based TN, opening opportunities for future work.

Humans are adept at leveraging visual cues from lip movements for recognizing speech in adverse listening conditions. Audio-Visual Speech Recognition (AVSR) models follow similar approach to achieve robust speech recognition in noisy conditions. In this work, we present a multilingual AVSR model incorporating several enhancements to improve performance and audio noise robustness. Notably, we adapt the recently proposed Fast Conformer model to process both audio and visual modalities using a novel hybrid CTC/RNN-T architecture. We increase the amount of audio-visual training data for six distinct languages, generating automatic transcriptions of unlabelled multilingual datasets (VoxCeleb2 and AVSpeech). Our proposed model achieves new state-of-the-art performance on the LRS3 dataset, reaching WER of 0.8%. On the recently introduced MuAViC benchmark, our model yields an absolute average-WER reduction of 11.9% in comparison to the original baseline. Finally, we demonstrate the ability of the proposed model to perform audio-only, visual-only, and audio-visual speech recognition at test time.

We present a novel Speech Augmented Language Model (SALM) with multitask and in-context learning capabilities. SALM comprises a frozen text LLM, a audio encoder, a modality adapter module, and LoRA layers to accommodate speech input and associated task instructions. The unified SALM not only achieves performance on par with task-specific Conformer baselines for Automatic Speech Recognition (ASR) and Speech Translation (AST), but also exhibits zero-shot in-context learning capabilities, demonstrated through keyword-boosting task for ASR and AST. Moreover, speech supervised in-context training is proposed to bridge the gap between LLM training and downstream speech tasks, which further boosts the in-context learning ability of speech-to-text models. Proposed model is open-sourced via NeMo toolkit1.

This paper presents an overview and evaluation of some of the end-to-end ASR models on long-form audio. We study three categories of Automatic Speech Recognition(ASR) models based on their core architecture: (1) convolutional, (2) convolutional with squeeze-and-excitation, and (3) convolutional models with attention. We selected one ASR model from each category and evaluated the Word Error Rate, maximum audio length and real-time factor for each model on a variety of long audio benchmarks: Earnings-21 and 22, CORAAL, and TED-LIUM3. The model from the category of self-attention with local attention and global token has the best accuracy compared to other architectures. We also compared models with CTC and RNNT decoders and showed that CTC-based models are more robust and efficient than RNNT on long form audio.

In this paper, we propose an efficient and accurate streaming speech recognition model based on the FastConformer architecture. We adapted the FastConformer architecture for streaming applications through: (1) constraining both the look-ahead and past contexts in the encoder, and (2) introducing an activation caching mechanism to enable the non-autoregressive encoder to operate autoregressively during inference. The proposed model is thoughtfully designed in a way to eliminate the accuracy disparity between the train and inference time which is common for many streaming models. Furthermore, our proposed encoder works with various decoder configurations including Connectionist Temporal Classification (CTC) and RNN-Transducer (RNNT) decoders. We evaluate the proposed model and demonstrate that it can achieve better accuracy with lower latency and inference time compared to a conventional buffered streaming model baseline.

Discrete audio representation, aka audio tokenization, has seen renewed interest driven by its potential to facilitate the application of text language modeling approaches in audio domain. To this end, various compression and representation-learning based tokenization schemes have been proposed. However, there is limited investigation into the performance of compression-based audio tokens compared to well-established mel-spectrogram features across various speaker and speech related tasks. In this paper, we evaluate compression based audio tokens on three tasks: Speaker Verification, Diarization and (Multi-lingual) Speech Recognition. Our findings indicate that (i) the models trained on audio tokens perform competitively, on average within 1% of mel-spectrogram features for all the tasks considered, and do not surpass them yet. (ii) these models exhibit robustness for out-of-domain narrowband data, particularly in speaker tasks. (iii) audio tokens allow for compression to 20x compared to mel-spectrogram features with minimal loss of performance in speech and speaker related tasks, which is crucial for low bit-rate applications, and (iv) the examined Residual Vector Quantization (RVQ) based audio tokenizer exhibits a low-pass frequency response characteristic, offering a plausible explanation for the observed results, and providing insight for future tokenizer designs.

This paper proposes Transducers with Pronunciation-aware Embeddings (PET). Unlike conventional Transducers where the decoder embeddings for different tokens are trained independently, the PET model’s decoder embedding incorporates shared components for text tokens with the same or similar pronunciations. With experiments conducted in multiple datasets in Mandarin Chinese and Korean, we show that PET models consistently improve speech recognition accuracy compared to conventional Transducers. Our investigation also uncovers a phenomenon that we call error chain reactions. Instead of recognition errors being evenly spread throughout an utterance, they tend to group together, with subsequent errors often following earlier ones. Our analysis shows that PET models effectively mitigate this issue by substantially reducing the likelihood of the model generating additional errors following a prior one. Our implementation will be open-sourced with the NeMo toolkit.

We propose SelfVC, a training strategy to iteratively improve a voice conversion model with self-synthesized examples. Previous efforts on voice conversion focus on factorizing speech into explicitly disentangled representations that separately encode speaker characteristics and linguistic content. However, disentangling speech representations to capture such attributes using task-specific loss terms can lead to information loss. In this work, instead of explicitly disentangling attributes with loss terms, we present a framework to train a controllable voice conversion model on entangled speech representations derived from self-supervised learning (SSL) and speaker verification models. First, we develop techniques to derive prosodic information from the audio signal and SSL representations to train predictive submodules in the synthesis model. Next, we propose a training strategy to iteratively improve the synthesis model for voice conversion, by creating a challenging training objective using self-synthesized examples. We demonstrate that incorporating such self-synthesized examples during training improves the speaker similarity of generated speech as compared to a baseline voice conversion model trained solely on heuristically perturbed inputs. Our framework is trained without any text and achieves state-of-the-art results in zero-shot voice conversion on metrics evaluating naturalness, speaker similarity, and intelligibility of synthesized audio.

Large Language Model (LLM) based text-to-speech (TTS) systems have demonstrated remarkable capabilities in handling large speech datasets and generating natural speech for new speakers.However, LLM-based TTS models are not robust as the generated output can contain repeating words, missing words and mis-aligned speech (referred to as hallucinations or attention errors), especially when the text contains multiple occurrences of the same token.We examine these challenges in an encoder-decoder transformer model and find that certain cross-attention heads in such models implicitly learn the text and speech alignment when trained for predicting speech tokens for a given text.To make the alignment more robust, we propose techniques utilizing CTC loss and attention priors that encourage monotonic cross-attention over the text tokens.Our guided attention training technique does not introduce any new learnable parameters and significantly improves robustness of LLM-based TTS models.

This paper introduces a highly efficient greedy decoding algorithm for Transducer-based speech recognition models. We redesign the standard nested-loop design for RNN-T decoding, swapping loops over frames and labels: the outer loop iterates over labels, while the inner loop iterates over frames searching for the next non-blank symbol. Additionally, we represent partial hypotheses in a special structure using CUDA tensors, supporting parallelized hypotheses manipulations. Experiments show that the label-looping algorithm is up to 2.0X faster than conventional batched decoding when using batch size 32. It can be further combined with other compiler or GPU call-related techniques to achieve even more speedup. Our algorithm is general-purpose and can work with both conventional Transducers and Token-and-Duration Transducers. We open-source our implementation to benefit the research community.

Incorporating speech understanding capabilities into pretrained large-language models has become a vital research direction (SpeechLLM). The previous architectures can be categorized as: i) GPT-style, prepend speech prompts to the text prompts as a sequence of LLM inputs like a decoder-only model; ii) T5-style, introduce speech cross-attention to each layer of the pretrained LLMs. We propose BESTOW architecture to bring the BESt features from $T w O$ Worlds into a single model that is highly efficient and has strong multitask capabilities. Moreover, there is no clear streaming solution for either style, especially considering the solution should generalize to speech multitask. We reformulate streamable SpeechLLM as a read-write policy problem and unifies the offline and streaming research with BESTOW architecture. Hence we demonstrate the first open-source SpeechLLM solution that enables Streaming and Multitask at scale (beyond ASR) at the same time. This streamable solution achieves very strong performance on a wide range of speech tasks (ASR, AST, SQA, unseen DynamicSuperb). It is end-to-end optimizable, with lower training/inference cost, and demonstrates LLM knowledge transferability to speech.

We introduce romanization encoding for script-heavy languages to optimize multilingual and code-switching Automatic Speech Recognition (ASR) systems. By adopting romanization encoding alongside a balanced concatenated tokenizer within a FastConformer-RNNT framework equipped with a Roman2Char module, we significantly reduce vocabulary and output dimensions, enabling larger training batches and reduced memory consumption. Our method decouples acoustic modeling and language modeling, enhancing the flexibility and adaptability of the system. In our study, applying this method to Mandarin-English ASR resulted in a remarkable 63.51% vocabulary reduction and notable performance gains of 13.72% and 15.03% on SEAME code-switching benchmarks. Ablation studies on MandarinKorean and Mandarin-Japanese highlight our method’s strong capability to address the complexities of other script-heavy languages, paving the way for more versatile and effective multilingual ASR systems.

This paper presents a new method for training sequence-to-sequence models for speech recognition and translation tasks. Instead of the traditional approach of training models on short segments containing only lowercase or partial punctuation and capitalization (PnC) sentences, we propose training on longer utterances that include complete sentences with proper punctuation and capitalization. We achieve this by using the FastConformer architecture which allows training 1 Billion parameter models with sequences up to 60 seconds long with full attention. However, while training with PnC enhances the overall performance, we observed that accuracy plateaus when training on sequences longer than 40 seconds across various evaluation settings. Our proposed method significantly improves punctuation and capitalization accuracy, showing a 25% relative word error rate (WER) improvement on the Earnings-21 and Earnings-22 benchmarks. Additionally, training on longer audio segments increases the overall model accuracy across speech recognition and translation benchmarks. The model weights and training code are open-sourced though NVIDIA NeMo.121https://github.com/NVIDIA/NeMo2https://huggingface.co/nvidia/parakeet-tdt_ctc-1.1b

Speech foundation models have achieved state-of-the-art (SoTA) performance across various tasks, such as automatic speech recognition (ASR) in hundreds of languages. However, multi-speaker ASR remains a challenging task for these models due to data scarcity and sparsity. In this paper, we present approaches to enable speech foundation models to process and understand multi-speaker speech with limited training data. Specifically, we adapt a speech foundation model for the multi-speaker ASR task using only telephonic data. Remarkably, the adapted model also performs well on meeting data without any fine-tuning, demonstrating the generalization ability of our approach. We conduct several ablation studies to analyze the impact of different parameters and strategies on model performance. Our findings highlight the effectiveness of our methods. Results show that less parameters give better overall cpWER, which, although counterintuitive, provides insights into adapting speech foundation models for multi-speaker ASR tasks with minimal annotated data.

Given recent advances in generative AI technology, a key question is how large language models (LLMs) can enhance acoustic modeling tasks using text decoding results from a frozen, pretrained automatic speech recognition (ASR) model. To explore new capabilities in language modeling for speech processing, we introduce the generative speech transcription error correction (GenSEC) challenge. This challenge comprises three post-ASR language modeling tasks: (i) post-ASR transcription correction, (ii) speaker tagging, and (iii) emotion recognition. These tasks aim to emulate future LLM-based agents handling voice-based interfaces while remaining accessible to a broad audience by utilizing open pretrained language models or agent-based APIs. We also discuss insights from baseline evaluations, as well as lessons learned for designing future evaluations.

Traditional automatic speech recognition (ASR) models output lower-cased words without punctuation marks, which reduces readability and necessitates a subsequent text processing model to convert ASR transcripts into a proper format. Simultaneously, the development of end-to-end ASR models capable of predicting punctuation and capitalization presents several challenges, primarily due to limited data availability and shortcomings in the existing evaluation methods, such as inadequate assessment of punctuation prediction. In this paper, we introduce a LibriSpeech-PC benchmark designed to assess the punctuation and capitalization prediction capabilities of end-to-end ASR models. The benchmark includes a LibriSpeech-PC dataset with restored punctuation and capitalization, a novel evaluation metric called Punctuation Error Rate (PER) that focuses on punctuation marks, and initial baseline models. All code, data, and models are publicly available.

Conformer-based models have become the dominant end-to-end architecture for speech processing tasks. With the objective of enhancing the conformer architecture for efficient training and inference, we carefully redesigned Conformer with a novel downsampling schema. The proposed model, named Fast Conformer(FC), is 2.8 × faster than the original Conformer, supports scaling to Billion parameters without any changes to the core architecture and also achieves state-of-the-art accuracy on Automatic Speech Recognition benchmarks. To enable transcription of long-form speech up to 11 hours, we replaced global attention with limited context attention post-training, while also improving accuracy through fine-tuning with the addition of a global token. Fast Conformer, when combined with a Transformer decoder also outperforms the original Conformer in accuracy and in speed for Speech Translation and Spoken Language Understanding.

We introduce VANI, a very lightweight multi-lingual accent controllable speech synthesis system. Our model builds upon disentanglement strategies proposed in RADMMM[1] and supports explicit control of accent, language, speaker and fine-grained F0and energy features for speech synthesis. We utilize the Indic languages dataset, released for LIMMITS 2023 as part of ICASSP Signal Processing Grand Challenge, to synthesize speech in 3 different languages. Our model supports transferring the language of a speaker while retaining their voice and the native accent of the target language. We utilize the large-parameter RADMMM model for Track 1 and lightweight VANI model for Track 2 and 3 of the competition.

In this paper, we extend previous self-supervised approaches for language identification by experimenting with Conformer based architecture in a multilingual pre-training paradigm. We find that pre-trained speech models optimally encode language discriminatory information in lower layers. Further, we demonstrate that the embeddings obtained from these layers are significantly robust to classify unseen languages and different acoustic environments without additional training. After fine-tuning a pre-trained Conformer model on the VoxLin-gua107 dataset, we achieve results similar to current state-of-the-art systems for language identification. More, our model accomplishes this with 5x less parameters. We open-source the model through the NVIDIA NeMo toolkit.

In this work, we propose a zero-shot voice conversion method using speech representations trained with self-supervised learning. First, we develop a multi-task model to decompose a speech utterance into features such as linguistic content, speaker characteristics, and speaking style. To disentangle content and speaker representations, we propose a training strategy based on Siamese networks that encourages similarity between the content representations of the original and pitch-shifted audio. Next, we develop a synthesis model with pitch and duration predictors that can effectively reconstruct the speech signal from its decomposed representation. Our framework allows controllable and speaker-adaptive synthesis to perform zero-shot any-to-any voice conversion achieving state-of-the-art results on metrics evaluating speaker similarity, intelligibility, and naturalness. Using just 10 seconds of data for a target speaker, our framework can perform voice swapping and achieves a speaker verification EER of 5.5% for seen speakers and 8.4% for unseen speakers.1

This paper presents a framework based on Weighted Finite-State Transducers (WFST) to simplify the development of modifications for RNN-Transducer (RNN-T) loss. Existing implementations of RNN-T use CUDA-related code, which is hard to extend and debug. WFSTs are easy to construct and extend, and allow debugging through visualization. We introduce two WFST-powered RNN-T implementations: (1) "Compose-Transducer", based on a composition of the WFST graphs from acoustic and textual schema – computationally competitive and easy to modify; (2) "Grid-Transducer", which constructs the lattice directly for further computations – most compact, and computationally efficient. We illustrate the ease of extensibility through introduction of a new W-Transducer loss – the adaptation of the Connectionist Temporal Classification with Wild Cards. W-Transducer (W-RNNT) consistently outperforms the standard RNN-T in a weakly-supervised data setup with missing parts of transcriptions at the beginning and end of utterances. All RNN-T losses are implemented with the k2 framework and are available in the NeMo toolkit.

This paper proposes a modification to RNN-Transducer (RNN-T) models for automatic speech recognition (ASR). In standard RNN-T, the emission of a blank symbol consumes exactly one input frame; in our proposed method, we introduce additional blank symbols, which consume two or more input frames when emitted. We refer to the added symbols as big blanks, and the method multi-blank RNN-T. For training multi-blank RNN-Ts, we propose a novel logit under-normalization method in order to prioritize emissions of big blanks. With experiments on multiple languages and datasets, we show that multi-blank RNN-T methods could bring relative speedups of over +90%/+139% to model inference for English Librispeech and German Multilingual Librispeech datasets, respectively. The multi-blank RNN-T method also improves ASR accuracy consistently. We will release our implementation of the method in the NeMo (https://github.com/NVIDIA/NeMo) toolkit.

We propose CONF-TSASR, a non-autoregressive end-to-end time-frequency domain architecture for single-channel target-speaker automatic speech recognition (TS-ASR). The model consists of a TitaNet based speaker embedding module, a Conformer based masking as well as ASR modules. These modules are jointly optimized to transcribe a target-speaker, while ignoring speech from other speakers. For training we use Connectionist Temporal Classification (CTC) loss and introduce a scale-invariant spectrogram reconstruction loss to encourage the model better separate the target-speaker’s spectrogram from mixture. We obtain state-of-the-art target-speaker word error rate (TS-WER) on WSJ0-2mix-extr (4.2%). Further, we report for the first time TS-WER on WSJ0-3mix-extr (12.4%), LibriSpeech2Mix (4.2%) and LibriSpeech3Mix (7.6%) datasets, establishing new benchmarks for TS-ASR. The proposed model will be open-sourced through NVIDIA NeMo toolkit.

This paper introduces a novel Token-and-Duration Transducer (TDT) architecture for sequence-to-sequence tasks. TDT extends conventional RNN-Transducer architectures by jointly predicting both a token and its duration, i.e. the number of input frames covered by the emitted token. This is achieved by using a joint network with two outputs which are independently normalized to generate distributions over tokens and durations. During inference, TDT models can skip input frames guided by the predicted duration output, which makes them significantly faster than conventional Transducers which process the encoder output frame by frame. TDT models achieve both better accuracy and significantly faster inference than conventional Transducers on different sequence transduction tasks. TDT models for Speech Recognition achieve better accuracy and up to 2.82X faster inference than conventional Transducers. TDT models for Speech Translation achieve an absolute gain of over 1 BLEU on the MUST-C test compared with conventional Transducers, and its inference is 2.27X faster. In Speech Intent Classification and Slot Filling tasks, TDT models improve the intent accuracy by up to over 1% (absolute) over conventional Transducers, while running up to 1.28X faster. Our implementation of the TDT model will be open-sourced with the NeMo (https://github.com/NVIDIA/NeMo) toolkit.

We propose an end-to-end Automatic Speech Recognition (ASR) system that can be trained on transcribed speech data, text-only data, or a mixture of both.The proposed model uses an integrated auxiliary block for text-based training.This block combines a non-autoregressive multi-speaker text-to-melspectrogram generator with a GAN-based enhancer to improve the spectrogram quality.The proposed system can generate a mel-spectrogram dynamically during training.It can be used to adapt the ASR model to a new domain by using text-only data from this domain.We demonstrate that the proposed training method significantly improves ASR accuracy compared to the system trained on transcribed speech only.It also surpasses cascade TTS systems with the vocoder in the adaptation quality and training speed.

The number of end-to-end speech recognition models grows every year.These models are often adapted to new domains or languages resulting in a proliferation of expert systems that achieve great results on target data, while generally showing inferior performance outside of their domain of expertise.We explore combination of such experts via confidence-based ensembles: ensembles of models where only the output of the most-confident model is used.We assume that models' target data is not available except for a small validation set.We demonstrate effectiveness of our approach with two applications.First, we show that a confidence-based ensemble of 5 monolingual models outperforms a system where model selection is performed via a dedicated language identification block.Second, we demonstrate that it is possible to combine base and adapted models to achieve strong results on both original and target data.We validate all our results on multiple datasets and model architectures.

In this paper, we propose TitaNet, a novel neural network architecture for extracting speaker representations. We employ 1D depth-wise separable convolutions with Squeeze-and-Excitation (SE) layers with global context followed by channel attention based statistics pooling layer to map variable-length utterances to a fixed-length embedding (t-vector). TitaNet is a scalable architecture and achieves state-of-the-art performance on speaker verification task with an equal error rate (EER) of 0.68% on the VoxCeleb1 trial file and also on speaker diarization tasks with diarization error rate (DER) of 1.73% on AMI-MixHeadset, 1.99% on AMI-Lapel and 1.11% on CH109. Furthermore, we investigate various sizes of TitaNet and present a light TitaNet-S model with only 6M parameters that achieve near state-of-the-art results in diarization tasks.

This paper describes Mixer-TTS, a non-autoregressive model for mel-spectrogram generation. The model is based on the MLP-Mixer architecture adapted for speech synthesis. The basic Mixer-TTS contains pitch and duration predictors, with the latter being trained with an unsupervised TTS alignment framework. Alongside the basic model, we propose the extended version which additionally uses token embeddings from a pre-trained language model. Basic Mixer-TTS and its extended version achieve a mean opinion score (MOS) of 4.05 and 4.11, respectively, compared to a MOS of 4.27 of original LJSpeech samples. Both versions have a small number of parameters and enable much faster speech synthesis compared to the models with similar quality.

Inverse text normalization (ITN) is an essential post-processing step in automatic speech recognition (ASR).It converts numbers, dates, abbreviations, and other semiotic classes from the spoken form generated by ASR to their written forms.One can consider ITN as a Machine Translation task and use neural sequence-tosequence models to solve it.Unfortunately, such neural models are prone to hallucinations that could lead to unacceptable errors.To mitigate this issue, we propose a single-pass token classifier model that regards ITN as a tagging task.The model assigns a replacement fragment to every input token or marks it for deletion or copying without changes.We present a method of dataset preparation, based on granular alignment of ITN examples.The proposed model is less prone to hallucination errors.The model is trained on the Google Text Normalization dataset and achieves state-of-the-art sentence accuracy on both English and Russian test sets.One-to-one correspondence between tags and input words improves the interpretability of the model's predictions, simplifies debugging, and allows for post-processing corrections.The model is simpler than sequence-to-sequence models and easier to optimize in production settings.The model and the code to prepare the dataset is published as part of NeMo project 1 .

Text normalization (TN) systems in production are largely rule-based using weighted finite-state transducers (WFST).However, WFST-based systems struggle with ambiguous input when the normalized form is context-dependent.On the other hand, neural text normalization systems can take context into account but they suffer from unrecoverable errors and require labeled normalization datasets, which are hard to collect.We propose a new hybrid approach that combines the benefits of rule-based and neural systems.First, a non-deterministic WFST outputs all normalization candidates, and then a neural language model picks the best one -similar to shallow fusion for automatic speech recognition.While the WFST prevents unrecoverable errors, the language model resolves contextual ambiguity.The approach is easy to extend and we show it is effective.It achieves comparable or better results than existing state-of-theart TN models.

This paper presents novel Weighted Finite-State Transducer (WFST) topologies to implement Connectionist Temporal Classification (CTC)-like algorithms for automatic speech recognition. Three new CTC variants are proposed: (1) the "compact-CTC", in which direct transitions between units are replaced with back-off transitions; (2) the "minimal-CTC", that only adds self-loops when used in WFST-composition; and (3) the "selfless-CTC" variants, which disallows self-loop for non-blank units. Compact-CTC allows for 1.5 times smaller WFST decoding graphs and reduces memory consumption by two times when training CTC models with the LF-MMI objective without hurting the recognition accuracy. Minimal-CTC reduces graph size and memory consumption by two and four times for the cost of a small accuracy drop. Using selfless-CTC can improve the accuracy for wide context window models.

This paper presents a class of new fast non-trainable entropy-based confidence estimation methods for automatic speech recognition. We show how per-frame entropy values can be normalized and aggregated to obtain a confidence measure per unit and per word for Connectionist Temporal Classification (CTC) and Recurrent Neural Network Transducer (RNN-T) models. Proposed methods have similar computational complexity to the traditional method based on the maximum per-frame probability, but they are more adjustable, have a wider effective threshold range, and better push apart the confidence distributions of correct and incorrect words. We evaluate the proposed confidence measures on LibriSpeech test sets, and show that they are up to 2 and 4 times better than confidence estimation based on the maximum per-frame probability at detecting incorrect words for Conformer-CTC and Conformer-RNN-T models, respectively.

Automatic speech recognition models are often adapted to improve their accuracy in a new domain. A potential drawback of model adaptation to new domains is catastrophic forgetting, where the Word Error Rate on the original domain is significantly degraded. This paper addresses the situation when we want to simultaneously adapt automatic speech recognition models to a new domain and limit the degradation of accuracy on the original domain without access to the original training dataset. We propose several techniques such as a limited training strategy and regularized adapter modules for the Transducer encoder, prediction, and joiner network. We apply these methods to the Google Speech Commands and to the UK and Ireland English Dialect speech data set and obtain strong results on the new target domain while limiting the degradation on the original domain.

We present MarbleNet, an end-to-end neural network for Voice Activity Detection (VAD). MarbleNet is a deep residual network composed from blocks of 1D time-channel separable convolution, batch-normalization, ReLU and dropout layers. When compared to a state-of-the-art VAD model, MarbleNet is able to achieve similar performance with roughly 1/10-th the parameter cost. We further conduct extensive ablation studies on different training methods and choices of parameters in order to study the robustness of MarbleNet in real-world VAD tasks.

In this paper, we demonstrate the efficacy of transfer learning and continuous learning for various automatic speech recognition (ASR) tasks using end-to-end models trained with CTC loss. We start with a large pre-trained English ASR model and show that transfer learning can be effectively and easily performed on: (1) different English accents, (2) different languages (from English to German, Spanish, Russian, or from Mandarin to Cantonese) and (3) application-specific domains. Our extensive set of experiments demonstrate that in all three cases, transfer learning from a good base model has higher accuracy than a model trained from scratch. Our results indicate that, for fine-tuning, larger pre-trained models are better than small pre-trained models, even if the dataset for fine-tuning is small. We also show that transfer learning significantly speeds up convergence, which could result in significant cost savings when training with large datasets.

This paper introduces a new multi-speaker English dataset for training text-to-speech models.The dataset is based on Lib-riVox audiobooks and Project Gutenberg texts, both in the public domain.The new dataset contains about 292 hours of speech from 10 speakers with at least 17 hours per speaker sampled at 44.1 kHz.To select speech samples with high quality, we considered audio recordings with a signal bandwidth of at least 13 kHz and a signal-to-noise ratio (SNR) of at least 32 dB.The dataset is publicly released at "http://www.openslr.org/109/".

In the English speech-to-text (STT) machine learning task, acoustic models are conventionally trained on uncased Latin characters, and any necessary orthography (such as capitalization, punctuation, and denormalization of non-standard words) is imputed by separate post-processing models.This adds complexity and limits performance, as many formatting tasks benefit from semantic information present in the acoustic signal but absent in transcription.Here we propose a new STT task: endto-end neural transcription with fully formatted text for target labels.We present baseline Conformer-based models trained on a corpus of 5,000 hours of professionally transcribed earnings calls, achieving a CER of 1.7.As a contribution to the STT research community, we release the corpus free for noncommercial use. 1

Inverse text normalization (ITN) converts spoken-domain automatic speech recognition (ASR) output into written-domain text to improve the readability of the ASR output. Many state-of-the-art ITN systems use hand-written weighted finite-state transducer(WFST) grammars since this task has extremely low tolerance to unrecoverable errors. We introduce an open-source Python WFST-based library for ITN which enables a seamless path from development to production. We describe the specification of ITN grammar rules for English, but the library can be adapted for other languages. It can also be used for written-to-spoken text normalization. We evaluate the NeMo ITN library using a modified version of the Google Text normalization dataset.

In this work, we introduce a simple yet efficient post-processing model for automatic speech recognition. Our model has Transformer-based encoder-decoder architecture which "translates" acoustic model output into grammatically and semantically correct text. We investigate different strategies for regularizing and optimizing the model and show that extensive data augmentation and the initialization with pretrained weights are required to achieve good performance. On the LibriSpeech benchmark, our method demonstrates significant improvement in word error rate over the baseline acoustic model with greedy decoding, especially on much noisier dev-other and test-other portions of the evaluation dataset. Our model also outperforms baseline with 6-gram language model re-scoring and approaches the performance of re-scoring with Transformer-XL neural language model.

We propose a new end-to-end neural acoustic model for automatic speech recognition. The model is composed of multiple blocks with residual connections between them. Each block consists of one or more modules with 1D time-channel separable convolutional layers, batch normalization, and ReLU layers. It is trained with CTC loss. The proposed network achieves near state-of-the-art accuracy on LibriSpeech and Wall Street Journal, while having fewer parameters than all competing models. We also demonstrate that this model can be effectively fine-tuned on new datasets.

We present an MatchboxNet - an end-to-end neural network for speech command recognition. MatchboxNet is a deep residual network composed from blocks of 1D time-channel separable convolution, batch-normalization, ReLU and dropout layers. MatchboxNet reaches state-of-the-art accuracy on the Google Speech Commands dataset while having significantly fewer parameters than similar models. The small footprint of MatchboxNet makes it an attractive candidate for devices with limited computational resources. The model is highly scalable, so model accuracy can be improved with modest additional memory and compute. Finally, we show how intensive data augmentation using an auxiliary noise dataset improves robustness in the presence of background noise.

In this paper we report state-of-the-art results on LibriSpeech among end-to-end speech recognition models without any external training data.Our model, Jasper, uses only 1D convolutions, batch normalization, ReLU, dropout, and residual connections.To improve training, we further introduce a new layer-wise optimizer called NovoGrad.Through experiments, we demonstrate that the proposed deep architecture performs as well or better than more complex choices.Our deepest Jasper variant uses 54 convolutional layers.With this architecture, we achieve 2.95% WER using a beam-search decoder with an external neural language model and 3.86% WER with a greedy decoder on LibriSpeech test-clean.We also report competitive results on Wall Street Journal and the Hub5'00 conversational evaluation datasets.

We present SEBOOST, a technique for boosting the performance of existing stochastic optimization methods. SEBOOST applies a secondary optimization process in the subspace spanned by the last steps and descent directions. The method was inspired by the SESOP optimization method for large-scale problems, and has been adapted for the stochastic learning framework. It can be applied on top of any existing optimization method with no need to tweak the internal algorithm. We show that the method is able to boost the performance of different algorithms, and make them more robust to changes in their hyper-parameters. As the boosting steps of SEBOOST are applied between large sets of descent steps, the additional subspace optimization hardly increases the overall computational burden. We introduce two hyper-parameters that control the balance between the baseline method and the secondary optimization process. The method was evaluated on several deep learning tasks, demonstrating promising results.