This work learns shared word embeddings for acoustic and phonetic sequences.
problem Mapping variable-length acoustic and phonetic sequences to fixed-dimensional vectors.
method Weak supervision and binary classification task to predict word similarity.
result Best model achieves an F1 score of 0.95 for binary classification.
Automated vowel duration measurement improves on manual methods.
problem Manual vowel duration annotation is time-consuming and inconsistent.
method Structured prediction framework; acoustic feature functions; learning algorithm to minimize duration error.
result Trained model outperforms HMM-based aligner in automatic duration estimation.
System identifies language of transliterated text.
problem Users struggle to understand non-native language transliterated text.
method Feature extraction of phonetic syllables using LSTM network.
result System accurately identifies language of transliterated text.
Acoustic Neighbor Embeddings map speech and text to fixed dimensions for phonetic confusability.
problem Mapping speech and text to fixed dimensions for phonetic confusability.
method Adapting SNE to sequential inputs, training two encoder neural networks.
result More accurate results with low-dimensional embeddings in word recognition tasks.
Unsupervised learning of speech representations using WaveNet autoencoders.
problem Extract meaningful latent representations of speech signals.
method Applying autoencoding neural networks to speech waveforms, using a high capacity WaveNet decoder, and comparing three variants of latent representations.
result Comparable performance with top entries in the ZeroSpeech 2017 unsupervised acoustic unit discovery task.
Flexible SC framework converts voices from non-aligned corpora.
problem Limited practical applications of SC due to lack of parallel corpora.
method Variational auto-encoder framework for non-parallel corpora.
result Framework enables spectral conversion without parallel corpora or alignments.
User-specific KWS system learns new keywords on-device.
problem Out-of-vocabulary problem in traditional KWS systems.
method Query-by-example enrollment and testing, using phonetic posteriors and FST.
result Promising performance on two keywords, preserving simplicity.
Telephonetic improves robustness of neural language models to ASR errors.
problem Handling ASR errors and semantic noise in neural language models trained on written text.
method Data augmentation framework using character-level and embedding space perturbations.
result Achieves state-of-the-art perplexity of 37.49 on Penn Treebank corpus.
Neural model detects phoneme boundaries from speech, outperforming baselines.
problem Phoneme boundary detection for speech processing applications.
method Learnable segmental features with a structured loss function.
result Model achieves state-of-the-art performance on TIMIT and Buckeye corpora.
DGCCA learns nonlinear transformations for multiple data views.
problem Learning informative nonlinear transformations from multiple data views.
method Deep Generalized Canonical Correlation Analysis (DGCCA) combines deep learning and generalized CCA.
result DGCCA representations outperform existing methods in phonetic transcription and hashtag recommendation.
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.
Study combines speaker verification and voice trigger detection in a single network.
problem Separate training for speaker verification and voice trigger detection.
method Multi-task learning with a single network trained on both tasks.
result Single network achieves comparable accuracy to independent models for each task.
End-to-end system improves speaker verification using attention mechanism.
problem Improving text-dependent speaker verification accuracy.
method Speaker discriminative CNNs extract features, attention mechanism combines them, end-to-end training optimizes system.
result The proposed system achieves better performance on Windows 10 speaker verification task.
Hybrid ASR systems can model graphemes effectively using chenones, outperforming traditional methods.
problem Traditional hybrid ASR systems struggle with English's poor grapheme-phoneme correspondence.
method Leveraging tied context-dependent graphemes (chenones) to model graphemes directly.
result Chenone-based systems significantly outperform senone baselines by 4.5% to 11.1% on English datasets.
Improved voice trigger detection in noisy environments.
problem Complex acoustic environments and lack of trigger phrase training data.
method Two-stage cascaded architecture with multi-task learning.
result Model reduces errors by half compared to baseline in challenging conditions.
Improves TTS accuracy by correcting context-dependent units.
problem Improves text-to-speech accuracy through speaker adaptation.
method Statistical model predicting context-dependent phonetic unit classes and their mean error values.
result Corrected boundaries of units improve TTS accuracy compared to HMM segmentation.
WaveNet reconstructs speech from brain activity, revealing acoustic features.
problem Reconstructing speech from brain activity with limited data.
method WaveNet model applied to STG intracranial recordings.
result WaveNet models reveal phoneme-level acoustic features.
In some speaker recognition scenarios we find conversations recorded simultaneously over multiple channels. That is the case of the interviews in the NIST SRE dataset. To take advantage of that, we propose a modification of the PLDA model that considers two different inter-session variability terms. The first term is t…
Model natural speech using unsupervised deep VAEs.
problem Process unlabelled speech data efficiently.
method Convolutional Variational Autoencoder (cVAE) with latent space arithmetic.
result Modify speech phonetic content or speaker identity without parallel data.
Bayesian SHMM models speech units from unannotated speech.
problem Discover latent speech units from unannotated data.
method Bayesian Subspace Hidden Markov Model (SHMM).
result System performs well on synthesis quality, lower bitrate.
End-to-end model detects articulatory features from speech data.
problem Detecting articulatory features from speech data for various applications.
method Apply Listen, Attend and Spell (LAS) architecture and attention models.
result End-to-end training of manners and places of articulation detectors.
Cryptonite tests NLP models with cryptic crossword clues.
problem Ambiguity in language poses a challenge for NLP models.
method Cryptic crossword dataset based on cryptic clues.
result Fine-tuning T5-Large on Cryptonite achieves only 7.6% accuracy.
AeGAN improves speech clarity in noisy environments.
problem Improving speech recognition in crowded noisy environments.
method Generative adversarial networks (GAN) with a novel architecture.
result The proposed framework outperforms traditional and learning-based methods.
Long Short-Term Memory (LSTM) is a recurrent neural network (RNN) architecture that has been designed to address the vanishing and exploding gradient problems of conventional RNNs. Unlike feedforward neural networks, RNNs have cyclic connections making them powerful for modeling sequences. They have been successfully u…
Deep learning converts one singer's voice to another without supervision.
problem Unsupervised singing voice conversion.
method Deep learning network with a single CNN encoder, WaveNet decoder, and classifier.
result Natural, recognizable singing voices converted without supervision.
Model learns accent patterns from small datasets.
problem Automatic speech recognition struggles with rare accents.
method Automatically retrieves phonological generalizations from a small dataset.
result Model generates a million phonological variations of words.
The EM training algorithm of the classical i-vector extractor is often incorrectly described as a maximum-likelihood method. The i-vector model is however intractable: the likelihood itself and the hidden-variable posteriors needed for the EM algorithm cannot be computed in closed form. We show here that the classical …
Improves cross-lingual NER by projecting entities from one language to another.
problem Limited annotated corpora for named entity recognition in many languages.
method Uses machine translation twice: first for sentences, then for entities; matches based on ortho- and phonetic similarity; identifies matches using distributional statistics.
result Improves cross-lingual NER by an average of 4.1 points on 5 diverse languages.
mGRN improves multivariate time series prediction by managing marginal and joint memories.
problem Extracting dependencies in multivariate sequential data with strong serial and cross-sectional dependencies.
method Developed a novel recurrent network architecture, Memory-Gated Recurrent Networks (mGRN), with gates for marginal and joint memories.
result mGRN consistently outperforms state-of-the-art architectures on various public datasets.
Paper proposes singing voice conversion without parallel data.
problem Convert singing voices without parallel data.
method Phonetic posterior feature, DBLSTM, vocoder.
result Successfully converts singing voices without parallel data.
Self-supervised model detects phoneme boundaries without annotations.
problem Unsupervised phoneme segmentation without manual annotations.
method Convolutional neural network trained with Noise-Contrastive Estimation.
result Model outperforms baselines on TIMIT and Buckeye corpora.
Improved acoustic word embeddings using shared decoder in multi-view encoders.
problem Learning discriminative acoustic word embeddings from text labels.
method Combining Siamese multi-view encoders with a shared decoder network to maximize the relationship between acoustic and text embeddings.
result 11.1% relative improvement in average precision on acoustic word discrimination task with WSJ dataset.
End-to-end speaker verification framework reduces text dependency.
problem Improving text-independent speaker verification.
method Jointly trains SE and ASR networks with triplet loss and adversarial gradient.
result Lower equal error rate and better text-independency compared to other approaches.
We present a systematic analysis on the performance of a phonetic recogniser when the window of input features is not symmetric with respect to the current frame. The recogniser is based on Context Dependent Deep Neural Networks (CD-DNNs) and Hidden Markov Models (HMMs). The objective is to reduce the latency of the sy…
Author name disambiguation in bibliographic databases is the problem of grouping together scientific publications written by the same person, accounting for potential homonyms and/or synonyms. Among solutions to this problem, digital libraries are increasingly offering tools for authors to manually curate their publica…
End-to-end TTS learns context features from text input.
problem Lack of understanding of context features learned by end-to-end TTS.
method Evaluated encoder outputs against context criteria derived from parametric TTS.
result Encoder outputs reflect linguistic and phonetic context features.
Improved acoustic modeling with attentive adversarial learning.
problem Domain variability in acoustic modeling.
method Proposes an attentive ADIT method with an attention mechanism to improve domain-invariance of deep features.
result Improves deep feature domain-invariance and senone-discriminativity over ADIT.
Improves speech separation by integrating time and frequency domains.
problem Speech separation using deep learning techniques.
method Proposes a framework that combines time and frequency domain features, using an embedding network and clustering.
result Obtained state-of-the-art results on WSJ0-2mix dataset.
Proposes a new neural network for text-dependent speaker verification.
problem Improves speaker verification by encoding phrase and speaker information.
method Uses differentiable alignment models to produce supervectors from utterances.
result Achieves competitive performance in text-dependent speaker verification tasks.
Paper proposes a deep learning model for understanding e-commerce addresses.
problem Challenges in parsing shipping addresses with no fixed format.
method Combines NLP techniques with pre-processing steps for addresses, uses RoBERTa for vector representations.
result RoBERTa model achieves 90% accuracy in sub-region classification for North and South Indian cities.
New framework learns phoneme metric from perception data.
problem Learn metric for phoneme similarity from data.
method Learning algorithms to derive metric from behavioral data.
result Framework outperforms previous metrics in phoneme prediction.