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.

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.

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.

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.

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.

Large language models (LLMs) have shown great promise for capturing contextual information in natural language processing tasks. We propose a novel approach to speaker diarization that incorporates the prowess of LLMs to exploit contextual cues in human dialogues. Our method builds upon an acoustic-based speaker diarization system by adding lexical information from an LLM in the inference stage. We model the multi-modal decoding process probabilistically and perform joint acoustic and lexical beam searches to incorporate cues from both modalities: audio and text. Our experiments demonstrate that infusing lexical knowledge from the LLM into an acoustics-only diarization system improves the overall speaker-attributed word error rate (SA-WER). The experimental results show that LLMs can provide complementary information to acoustic models for the speaker diarization task via the proposed beam search decoding approach showing up to 39.8% relative delta-SA-WER improvement from the baseline system. Thus, we substantiate that the proposed technique is able to exploit contextual information that is inaccessible to acoustics-only systems which is represented by speaker embeddings. In addition, these findings point to the potential of using LLMs to improve speaker diarization and other speech-processing tasks by capturing semantic and contextual cues.

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.

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.

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

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.

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.

Computational modeling of naturalistic conversations in clinical applications has seen growing interest in the past decade. An important use-case involves child-adult interactions within the autism diagnosis and intervention domain. In this paper, we address a specific sub-problem of speaker diarization, namely child-adult speaker classification in such dyadic conversations with specified roles. Training a speaker classification system robust to speaker and channel conditions is challenging due to inherent variability in the speech within children and the adult interlocutors. In this work, we propose the use of meta-learning, in particular prototypical networks which optimize a metric space across multiple tasks. By modeling every child-adult pair in the training set as a separate task during meta-training, we learn a representation with improved generalizability compared to conventional supervised learning. We demonstrate improvements over state-of-the-art speaker embeddings (x-vectors) under two evaluation settings: weakly supervised classification (upto 14.53% relative improvement in F1-scores) and clustering (upto relative 9.66% improvement in cluster purity). Our results show that protonets can potentially extract robust speaker embeddings for child-adult classification from speech.

Bird sound detection from real-field recordings is essential for identifying bird species in bioacoustic monitoring. Variations in the recording devices, environmental conditions, and the presence of vocalizations from other animals make the bird sound detection very challenging. In order to overcome these challenges, we propose an unsupervised algorithm comprising two main stages. In the first stage, a spectrogram enhancement technique is proposed using a multiple window Savitzky-Golay (MWSG) filter. We show that the spectrogram estimate using MWSG filter is unbiased and has lower variance compared with its single window counterpart. It is known that bird sounds are highly structured in the time-frequency (T-F) plane. We exploit these cues of prominence of T-F activity in specific directions from the enhanced spectrogram, in the second stage of the proposed method, for bird sound detection. In this regard, we use a set of four moving average filters that when applied to the enhanced spectrogram, yield directional spectrograms that capture the direction specific information. We propose a thresholding scheme on the time varying energy profile computed from each of these directional spectrograms to obtain frame-level binary decisions of bird sound activity. These individual decisions are then combined to obtain the final decision. Experiments are performed with three different datasets, with varying recording and noise conditions. Frame level F-score is used as the evaluation metric for bird sound detection. We find that the proposed method, on average, achieves higher F-score ($10.24\%$ relative) compared to the best of the six baseline schemes considered in this work.