Automatic Speech Recognition (ASR) is widely used within consumer devices such as mobile phones. Recently, personalization or on-device model fine-tuning has shown that adaptation of ASR models towards target user speech improves their performance over rare words or accented speech. Despite these gains, fine-tuning on user data (target domain) risks the personalized model to forget knowledge about its original training distribution (source domain) i.e. catastrophic forgetting, leading to subpar general ASR performance. A simple and efficient approach to combat catastrophic forgetting is to measure forgetting via a validation set that represents the source domain distribution. However, such validation sets are large and impractical for mobile devices. Towards this, we propose a novel method to subsample a substantially large validation set into a smaller one while maintaining the ability to estimate forgetting. We demonstrate the efficacy of such a dataset in mitigating forgetting by utilizing it to dynamically determine the number of ideal fine-tuning epochs. When measuring the deviations in per user fine-tuning epochs against a 50x larger validation set (oracle), our method achieves a lower mean-absolute-error (3.39) compared to randomly selected subsets of the same size (3.78-8.65). Unlike random baselines, our method consistently tracks the oracle’s behaviour across three different forgetting thresholds.

Data augmentation (DA) is ubiquitously used in training of Automatic Speech Recognition (ASR) models. DA offers increased data variability, robustness and generalization against different acoustic distortions. Recently, personalization of ASR models on mobile devices has been shown to improve Word Error Rate (WER). This paper evaluates data augmentation in this context and proposes persoDA; a DA method driven by user’s data utilized to personalize ASR [1] –[3]. persoDA aims to augment training with data specifically tuned towards acoustic characteristics of the end-user, as opposed to standard augmentation based on Multi-Condition Training (MCT) that applies random reverberation and noises. Our evaluation with an ASR conformer-based baseline trained on Librispeech and per-sonalized for VOICES [4] shows that persoDA achieves a 13.9% relative WER reduction over using standard data augmentation (using random noise & reverberation). Furthermore, persoDA shows 16% to 20% faster convergence over MCT.

Automatic Speech Recognition (ASR) still face challenges when recognizing time-variant rare-phrases. Contextual biasing (CB) modules bias ASR model towards such contextually-relevant phrases. During training, a list of biasing phrases are selected from a large pool of phrases following a sampling strategy. In this work we firstly analyse different sampling strategies to provide insights into the training of CB for ASR with correlation plots between the bias embeddings among various training stages. Secondly, we introduce a neighbourhood attention (NA) that localizes self attention (SA) to the nearest neighbouring frames to further refine the CB output. The results show that this proposed approach provides on average a 25.84% relative WER improvement on LibriSpeech sets and rare-word evaluation compared to the baseline.

On-device Automatic Speech Recognition (ASR) models trained on speech data of a large population might underperform for individuals unseen during training. This is due to a domain shift between user data and the original training data, differed by user’s speaking characteristics and environmental acoustic conditions. ASR personalisation is a solution that aims to exploit user data to improve model robustness. The majority of ASR personalisation methods assume labelled user data for supervision. Personalisation without any labelled data is challenging due to limited data size and poor quality of recorded audio samples. This work addresses unsupervised personalisation by developing a novel consistency based training method via pseudo-labelling. Our method achieves a relative Word Error Rate Reduction (WERR) of 17.3% on unlabelled training data and 8.1% on held-out data compared to a pre-trained model, and outperforms the current state-of-the art methods.

This paper introduces a two-pass on-device automatic speech recognition (ASR) system, which is developed for commercialized devices. The first pass of the system is based on a causal Conformer-transducer model to generate partial results from the input audio stream. After processing an entire input utterance in the first pass, the candidates for the final result are rescored with a full-context attention model in the second pass. To minimize the computational overhead from rescoring, we compress the full-context model by applying knowledge distillation (KD). The total model size is reduced by 35% after KD with a 0.02% absolute loss in word error rate (WER). We also introduce decoding techniques to boost the accuracy on the test cases mismatched with the distribution of the training set. The techniques include on-device personal adaptation, spell correction and handling incorrectly segmented speech, which solve the critical issues for production-grade systems. The whole system including the two-pass end-to-end (E2E) model and a language model (LM) occupies 72MB in storage after 8-bit quantization. We demonstrate the entire system on mobile devices and report results on test sets collected from the production environment. The developed system achieves 5.65% WER which surpasses the baseline system with 39% relative WER improvement.

We propose an adapter based multi-domain Transformer based language model (LM) for Transformer ASR. The model consists of a big size common LM and small size adapters. The model can perform multi-domain adaptation with only the small size adapters and its related layers. The proposed model can reuse the full fine-tuned LM which is fine-tuned using all layers of an original model. The proposed LM can be expanded to new domains by adding about 2% of parameters for a first domain and 13% parameters for after second domain. The proposed model is also effective in reducing the model maintenance cost because it is possible to omit the costly and time-consuming common LM pre-training process. Using proposed adapter based approach, we observed that a general LM with adapter can outperform a dedicated music domain LM in terms of word error rate (WER).

In this paper, we present a new on-device automatic speech recognition (ASR) system based on monotonic chunk-wise attention (MoChA) models trained with large (> 10K hours) corpus. We attained around 90% of a word recognition rate for general domain mainly by using joint training of connectionist temporal classifier (CTC) and cross entropy (CE) losses, minimum word error rate (MWER) training, layer-wise pretraining and data augmentation methods. In addition, we compressed our models by more than 3.4 times smaller using an iterative hyper low-rank approximation (LRA) method while minimizing the degradation in recognition accuracy. The memory footprint was further reduced with 8-bit quantization to bring down the final model size to lower than 39 MB. For on-demand adaptation, we fused the MoChA models with statistical n-gram models, and we could achieve a relatively 36% improvement on average in word error rate (WER) for target domains including the general domain.