New training method for ASR models reduces word error rate.
problem Traditional ASR training methods do not directly optimize word error rate.
method Training attention-based models to minimize expected word error rate using N-best lists.
result Improved performance by up to 8.2% compared to baseline.
State-level minimum Bayes risk (sMBR) training has become the de facto standard for sequence-level training of speech recognition acoustic models. It has an elegant formulation using the expectation semiring, and gives large improvements in word error rate (WER) over models trained solely using cross-entropy (CE) or co…
New method improves ASR word confidence for diverse applications.
problem Mitigating ASR errors and improving word error rate.
method Heterogeneous Word Confusion Network (HWCN) with score calibration.
result Word sequence with best overall confidence is more accurate than 1-best result.
Improved speech recognition model reduces error rate from 9.2% to 5.6%.
problem Improving speech recognition accuracy for voice search tasks.
method Structural and optimization improvements to Listen, Attend, and Spell (LAS) model.
result Significant reduction in word error rate (WER) from 9.2% to 5.6% on voice search task.
Unified framework for geometric computation of minimum-area homotopy.
problem Computing the minimum homotopy area of a closed curve.
method Unified combinatorial word approach combining geometric and algebraic methods.
result Unified geometric proof and constructive algorithm for minimum area homotopy.
This paper reports the performances of shallow word-level convolutional neural networks (CNN), our earlier work (2015), on the eight datasets with relatively large training data that were used for testing the very deep character-level CNN in Conneau et al. (2016). Our findings are as follows. The shallow word-level CNN…
End-to-end ASR model combines word and character representation for improved performance.
problem Difficulty in training with word-level supervision due to sparsity of examples.
method Multi-task learning framework combining word and character representations.
result Improved word-error rate (WER) by interpolating between word-level and character-level models.
Enhances speech recognition in new environments by embedding noise and scaling training data.
problem Improving speech recognition in unseen noisy environments.
method Embedding noise from unseen environments and scaling training data to 16,784 environments.
result Reduced word error rate from 34.04% to 15.46% on enhanced speech.
Seq2seq ASR adapts to speakers, improving performance by 25%.
problem Speaker adaptation for seq2seq ASR systems to match conventional methods.
method Applied Kullback-Leibler divergence and Linear Hidden Network adaptation to seq2seq models.
result 25% relative word error rate improvement with seq2seq model adaptation.
Large speech dataset for commercial use with 9.98% word error rate.
problem Creating a diverse speech recognition dataset for commercial purposes.
method Internet search for licensed audio data with transcriptions, training model on the dataset.
result Model trained on dataset achieves 9.98% word error rate on Librispeech's test-clean test set.
A hybrid ASR system using conformer architecture improves word-error-rate and training speed.
problem Improving word-error-rate and training efficiency for hybrid ASR systems.
method Used conformer architecture, applied time downsampling, and transposed convolutions.
result Conformer-based hybrid model achieves competitive results and significantly outperforms BLSTM-based hybrid model.
Three LF training criteria improve neural network acoustic models without cross-entropy pre-training.
problem Improving purely sequence-trained neural network acoustic models.
method Comparison of three lattice-free discriminative training criteria (MMI, bMMI, sMBR) on LVCSR tasks.
result LF-bMMI models outperform plain LF-MMI models by 5% WER on Switchboard datasets.
Paper proposes deep neural networks for nonparametric regression from dependent data.
problem Nonparametric regression from strongly mixing observations.
method Minimum error entropy principle applied to deep neural networks.
result Deep neural networks achieve minimax optimal convergence rates for Gaussian errors.
Improved speech recognition with cumulative adaptation methods.
problem Robust speech recognition in varying environments and speakers.
method Used a bidirectional LSTM neural network and i-vectors for adaptation.
result Achieved 13% relative improvement in word error rate.
FMP improves personalized ASR models on private devices.
problem Limited accuracy and privacy in federated fine-tuning of ASR models.
method FMP estimates global and personalized marginal distributions and adjusts NNLM probabilities.
result FMP achieves modest WER reductions on ASR rescoring tasks.
Minimum-norm solutions generalize well in over-parametrized neural networks.
problem Generalization error in over-parametrized neural networks.
method Analyzing three models: random feature model, two-layer neural network, and residual network.
result Generalization error for minimum-norm solutions is comparable to Monte Carlo rate, up to logarithmic terms.
The study analyzes how covariance estimation errors affect the global minimum-variance portfolio under heavy-tailed distributions.
problem The impact of covariance estimation errors on the global minimum-variance portfolio under heavy-tailed distributions.
method Characterization of covariance-estimation error's effect on GMVP suboptimality, derivation of regret identity and bound, application to heavy-tailed returns.
result The decision geometry of GMVP regret is invariant to a (p-1)-dimensional projection of the error matrix, with invariance to the covariance-scale direction as an exact special case.
Paper introduces a new optimisation method combining NG and Hessian Free for sequence training.
problem Overfitting and inefficiency in training DNNs with sequence criteria.
method Combines Natural Gradient and Hessian Free methods for better path traversal on parameter manifold.
result Achieves larger reductions in WER and lower WER compared to NG and HF methods.
In this paper we study the consistency of an empirical minimum error entropy (MEE) algorithm in a regression setting. We introduce two types of consistency. The error entropy consistency, which requires the error entropy of the learned function to approximate the minimum error entropy, is shown to be always true if the…
Grapheme ASR improves with G2G model that corrects spelling errors.
problem Rare long-tail words in non-phonemic languages like English.
method Train G2G model on text-to-speech data to rewrite character sequences into phonetically consistent forms.
result Reduces Word Error Rate by 3% to 11% over a strong graphemic baseline.
TeLeS improves ASR confidence estimation by considering temporal alignment and lexical errors.
problem Inaccurate confidence scores from E2E ASR models, especially for overconfident predictions.
method Proposes TeLeS, a novel confidence score that considers temporal alignment and lexical errors, and uses shrinkage loss to handle data imbalance.
result TeLeS generalizes well across different languages and ASR models, leading to significant WER reduction.
This work compares and evaluates various sampling methods for neural language models.
problem Lack of systematic comparison and myths about sampling methods.
method Monte Carlo sampling, importance sampling, compensated partial summation, noise contrastive estimation.
result All sampling methods can perform equally well if posterior probabilities are corrected.
High-order RNNs reduce speech recognition errors.
problem Vanishing gradients in RNNs.
method High-order RNNs with multiple connections from previous time steps.
result HORNNs reduce WER by 4.2% and 6.3% over RNNs.
We empirically characterize the performance of discriminative and generative LSTM models for text classification. We find that although RNN-based generative models are more powerful than their bag-of-words ancestors (e.g., they account for conditional dependencies across words in a document), they have higher asymptoti…
Streaming ASR with transformer achieves low WER.
problem Real-time ASR with speech recognition.
method Time-restricted self-attention and triggered attention mechanisms.
result 2.8% and 7.2% WER for LibriSpeech test data.
Algorithm extracts deterministic PDFA from probabilistic models with improved performance.
problem Learning deterministic models from probabilistic ones with noise.
method Adapted L* algorithm for probabilistic settings, using conditional probabilities and local tolerance.
result Achieves better performance on WER and NDCG than spectral extraction of WFAs.
A deep clustering model learns to separate audio sources without supervision.
problem Training deep clustering models requires supervision, limiting their applicability.
method Proposes an unsupervised spatial clustering approach to train a deep clustering system.
result The deep clustering model achieves similar performance to a multi-channel teacher without supervision.
Improved TF-IDF for word relevance in health-care social media documents.
problem Determining word relevance in informal documents.
method Semantic Sensitive TF-IDF (STF-IDF) method.
result Decreased TF-IDF mean error rate by 50% to 13.7%.
The depth of a link measures the minimum height of a resolving tree for the link whose leaves are all unlinks. We show that the depth of the closure of a strictly positive braid word is the length of the word minus the number of distinct letters.
LSTM LMs improve speech recognition by 8% using lattice rescoring.
problem Efficiently integrating LSTM LMs into speech recognition systems.
method Lattice rescoring algorithms using LSTM LMs.
result Reduces word error rate (WER) by 8% relative to N-gram LMs.
New method transfers word embeddings from large to small datasets efficiently.
problem Challenges in learning word embeddings from new domains with limited data.
method Group-sparse matrix factorization for transfer learning.
result Efficiently learns domain-specific word embeddings with less data.
This paper improves ASR robustness by learning domain invariant features.
problem Robustness issues in ASR due to mismatched training and testing distributions.
method Factorized Hierarchical Variational Autoencoder (FHVAE) for unsupervised learning of domain invariant features.
result 41% and 27% absolute word error rate reductions on mismatched domains.
A new batch optimisation framework using Natural Gradient improves DNN acoustic models.
problem Optimizing DNN acoustic models for better word error rate approximation.
method Proposes a Natural Gradient (NG) approach to sequence training, correcting the gradient based on local curvature of KL-divergence.
result The NG method converges more quickly and can be applied to any sequence discriminative training criterion.
Graphical lasso models ASR utterance dependencies for consistent WER estimation.
problem Modeling dependent structure among ASR utterances for accurate significance analysis.
method Graphical lasso for dependency modeling, followed by blockwise bootstrap resampling.
result Statistically consistent variance estimator of WER under mild conditions.
The paper introduces a model to measure ASR fairness, addressing key issues.
problem Measuring fairness in ASR systems for different subgroups.
method Mixed-effects Poisson regression to control nuisance factors and handle unobserved heterogeneity.
result The method effectively addresses WER gaps among subgroups and is flexible for practical analyses.
Improved speech recognition with faster training and inference.
problem Training very deep CNNs for speech recognition is difficult.
method Proposed SNDCNN using SELU activations instead of RELU and shortcut connections/BN.
result Achieved similar or lower WER with faster training and inference.
System builds Somali ASR for UN humanitarian efforts.
problem Developing ASR for under-resourced Somali language.
method Acoustic model training with annotated speech, neural architectures, language model data augmentation, acoustic data perturbation.
result Best system achieved 53.75% word error rate.
Study various complexity measures of curves on surfaces.
problem Measuring complexity of curves on surfaces.
method Examines minimum intersections, lengths of words, and covering degrees.
result Established relationships between these complexity measures.
Improved multilingual speech recognition with low latency for nine Indic languages.
problem Imbalance in training data across languages and low latency in interactive applications.
method Conditioning on a language vector and training language-specific adapter layers.
result Lower word error rate than monolingual E2E models and conventional systems.
Recent work on end-to-end automatic speech recognition (ASR) has shown that the connectionist temporal classification (CTC) loss can be used to convert acoustics to phone or character sequences. Such systems are used with a dictionary and separately-trained Language Model (LM) to produce word sequences. However, they a…
Jointly training with policy learning improves speech recognition performance.
problem Mismatch between CTC objective function and word error rate metric.
method Joint training with maximum likelihood and policy gradient.
result Joint training improves performance by 4% to 13%.
Paper proposes low-rank gradient approximation to save memory for deep neural network training.
problem Memory limitation on mobile devices for deep neural network training.
method Approximating gradient matrices using low-rank parameterization.
result Reduces training memory by about 33.0% for Adam optimization and 4.5% relative lower word error rate on ASR personalization task.
The main aim of this paper is to provide an analysis of gradient descent (GD) algorithms with gradient errors that do not necessarily vanish, asymptotically. In particular, sufficient conditions are presented for both stability (almost sure boundedness of the iterates) and convergence of GD with bounded, (possibly) non…
Improved community detection in heterogeneous SBM with side information.
problem Misclassification in community detection with noisy labels.
method Optimal weighted message passing and minimum energy flow.
result Optimal weighting improves misclassification rate in heterogeneous SBM.
Blockwise bootstrap improves ASR performance testing for correlated data.
problem Testing reliability of WER improvements between ASR systems.
method Divide evaluation utterances into nonoverlapping blocks and resample these blocks.
result The variance estimator of absolute WER difference is consistent under mild conditions.
This paper closes the accuracy gap between A2W models and sub-word models using English conversational speech data.
problem Closing the accuracy gap between direct acoustics-to-word models and sub-word models.
method Training an A2W model with orders of magnitude more data, optimizing model initialization, training data order, and regularization.
result Achieved word error rates of 8.8%/13.9% on Hub5-2000 Switchboard/CallHome test sets.
Abstract Coxeter groups have growth rates that are Perron numbers.
problem Understanding growth rates of Coxeter groups.
method Defined a class of Coxeter groups, ∞--spanned, and analyzed their growth rates. result For ∞--spanned Coxeter groups, geodesic growth rate strictly dominates word growth rate and appears to be a Perron number. While the channel capacity reflects a theoretical upper bound on the achievable information transmission rate in the limit of infinitely many bits, it does not characterise the information transfer of a given encoding routine with finitely many bits. In this note, we characterise the quality of a code (i. e. a given en…