New method improves dialogue agents focusing on simple utterances.
problem Dialogue agents often focus on simple utterances and suboptimal policies.
method Tempered Policy Gradient (TPG) methods to improve dialogue performance.
result Significant improvements in dialogue performance, especially in producing convincing utterances.
BOFFIN TTS optimizes hyper-parameters for new speaker adaptation.
problem Fine-tuning a pre-trained TTS model for a new speaker with limited data.
method Bayesian optimization to efficiently find optimal hyper-parameters.
result Average 30% improvement in speaker similarity over standard techniques.
Modeling long-range context for multi-function utterances in dialogues.
problem Complex dependencies across dialogue turns in long utterances.
method Adapted Convolutional Recurrent Neural Network (CRNN) to model interactions between utterances.
result Significantly outperforms existing work on CDA recognition on a tech forum dataset.
Deep neural network improves speaker verification with short utterances.
problem Challenges in speaker recognition with short utterances.
method Proposes deep neural network mapping methods to improve i-vector performance.
result Deep neural network mapping significantly improves speaker verification performance with short utterances.
Proposes a method to learn speaker embeddings for variable duration utterances.
problem Mismatch between training and testing utterance durations degrades speaker verification performance.
method Sliding window segmentation, LSTM, attentive pooling, segment-level and utterance-level embeddings, similarity loss.
result Significant improvement in robustness for duration variant utterances.
Meta-learning framework improves short utterance speaker recognition.
problem Poor performance of existing models with short utterances.
method Prototypical Networks with support and query sets, enforcing classification against entire training set.
result Significant performance gains on VoxCeleb datasets.
Paper presents a Siamese network for identifying more convincing evidence.
problem Identifying the more convincing argument in discussions.
method Proposes a Siamese neural network architecture for labeling evidence as more convincing.
result The Siamese network outperforms baselines on a new challenging data set.
ConvMFiT classifies client utterances in counseling dialogues.
problem Classifying client utterances for better counseling outcomes.
method Fine-tuning a pre-trained conversation model on counselor-client dialogues.
result ConvMFiT outperforms state-of-the-art models in classification.
Improves speaker verification for variable-duration utterances using a feature pyramid module.
problem Improving robustness for variable-duration utterances in speaker verification.
method Integrates a feature pyramid module into multi-scale aggregation to enhance speaker-discriminative information from multiple layers.
result Improves performance for both short and long utterances compared to state-of-the-art approaches.
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.
Self multi-head attention improves speaker recognition for long utterances.
problem Speaker recognition for long speech segments using Deep Learning.
method Convolutional Neural Network (CNN) for short-term features, self multi-head attention for long-term embeddings.
result Self multi-head attention outperforms other pooling methods by 18% relative EER on VoxCeleb1 dataset.
Improved far-field speaker verification for short utterances in noisy conditions.
problem Challenges in speaker verification on short utterances in uncontrolled noisy environments.
method Used deep neural network architectures (TDNN and ResNet) and experimented with various embedding extractors and training procedures.
result ResNet architectures outperform x-vector approach in speaker verification quality for both long and short utterances.
Simple baselines improve multimodal utterance learning.
problem Learning rich multimodal utterance representations.
method Conditional factorization of utterances into unimodal factors; extending to bimodal and trimodal factors.
result Optimal embeddings can be derived in closed form.
DGP learns speech recognition by modeling complex relationships between utterances.
problem Modeling complex relationships in speech recognition without relational data.
method Bayesian nonparametric deep learning method (DGP) that generates infinite probabilistic graphs.
result DGP successfully infers relationships among utterances without relational data during training.
A new type of End-to-End system for text-dependent speaker verification is presented in this paper. Previously, using the phonetically discriminative/speaker discriminative DNNs as feature extractors for speaker verification has shown promising results. The extracted frame-level (DNN bottleneck, posterior or d-vector) …
Improved voice conversion with semi-supervised learning.
problem Voice conversion with limited parallel data.
method Amortized variational inference with parallel and non-parallel utterances.
result Semi-supervised training improves voice conversion performance.
Recurrent sequence generators conditioned on input data through an attention mechanism have recently shown very good performance on a range of tasks in- cluding machine translation, handwriting synthesis and image caption gen- eration. We extend the attention-mechanism with features needed for speech recognition. We sh…
Speech enhancement model is used to map a noisy speech to a clean speech. In the training stage, an objective function is often adopted to optimize the model parameters. However, in most studies, there is an inconsistency between the model optimization criterion and the evaluation criterion on the enhanced speech. For …
Activates speech DNNs to generate understandable examples.
problem Difficulty in understanding DNN classifications for speech.
method Activation maximization to generate speech samples.
result Activation maximization can generate understandable speech samples.
Paper explores unsupervised transfer learning for SLU, improving model performance with unlabeled data.
problem Improving SLU model performance with limited labeled data.
method Uses ELMo embeddings for unsupervised pre-training and ELMo-Light for faster pre-training. Combines unsupervised and supervised transfer techniques.
result Unsupervised pre-training on unlabeled data significantly improves SLU performance, even outperforming conventional supervised transfer.
This paper presents sampling-based speech parameter generation using moment-matching networks for Deep Neural Network (DNN)-based speech synthesis. Although people never produce exactly the same speech even if we try to express the same linguistic and para-linguistic information, typical statistical speech synthesis pr…
The paper extracts structured data from physician-patient conversations, reducing clerical burden.
problem Mining insights from physician-patient conversations for electronic health record documentation.
method Created a dataset of transcripts and summaries, extracted noteworthy utterances, and improved model performance.
result Extracting noteworthy utterances significantly boosts model performance for recognizing diagnoses and RoS abnormalities.
Developers of text-to-speech synthesizers (TTS) often make use of human raters to assess the quality of synthesized speech. We demonstrate that we can model human raters' mean opinion scores (MOS) of synthesized speech using a deep recurrent neural network whose inputs consist solely of a raw waveform. Our best models …
Two large medical dialogue datasets for improving healthcare.
problem Improving healthcare through better dialogue systems.
method Building two large-scale medical dialogue datasets: MedDialog-EN and MedDialog-CN.
result The datasets are the largest medical dialogue datasets to date.
The traditional Sznajd model, as well as its Ochrombel simplification for opinion spreading, are applied to marketing with the help of advertising. The larger the lattice is the smaller is the amount of advertising needed to convince the whole market
Anonymizes speech data to protect privacy.
problem Protecting personal speech data from misuse.
method Extracts features, uses x-vectors, and neural models to synthesize anonymized speech.
result Effective in concealing speaker identities without compromising speech quality.
Discretizes special surfaces using Koenigs nets.
problem Integrable structure of special surfaces.
method Discretisation via Koenigs nets.
result Preserves integrable structure in discretization.
Inspired by recent work on neural network image generation which rely on backpropagation towards the network inputs, we present a proof-of-concept system for speech texture synthesis and voice conversion based on two mechanisms: approximate inversion of the representation learned by a speech recognition neural network,…
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.
Self-supervised method detects replay spoofing using acoustic configurations.
problem Challenges in collecting large-scale datasets for replay spoofing detection.
method Self-supervised pretraining of acoustic configurations using existing datasets.
result The method outperforms baseline by 30% on ASVspoof 2019 physical access dataset.
STOVE predicts convincing physical behavior in videos.
problem Learning models from videos with objects and interactions.
method Compositional state-space model combining image and dynamics models.
result Predicts videos with convincing physical behavior over hundreds of timesteps.
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.
A standard recipe for spoken language recognition is to apply a Gaussian back-end to i-vectors. This ignores the uncertainty in the i-vector extraction, which could be important especially for short utterances. A recent paper by Cumani, Plchot and Fer proposes a solution to propagate that uncertainty into the backend. …
Paper introduces methods to automatically generate SOAP notes from patient-physician conversations.
problem Burden of creating digital SOAP notes by physicians.
method Cluster2Sent algorithm for summarizing patient-physician conversations.
result Cluster2Sent algorithm outperforms existing methods by 8 ROUGE-1 points.
Study combines chit-chat and goal-oriented dialogue in fantasy games.
problem Combining naturalistic chit-chat with goal-oriented tasks in fantasy games.
method Trained a goal-oriented model with reinforcement learning against an imitation-learned chit-chat model using two approaches.
result Both models outperform a baseline and can converse naturally to achieve goals.
This paper develops a theory for group Lasso using a concept called strong group sparsity. Our result shows that group Lasso is superior to standard Lasso for strongly group-sparse signals. This provides a convincing theoretical justification for using group sparse regularization when the underlying group structure is …
This paper improves speech synthesis using a DGP with SRU for naturalness.
problem Improving naturalness in synthetic speech.
method Deep Gaussian process with a recurrent architecture using SRU.
result SRU-DGP outperforms other models in naturalness of synthetic speech.
GE2E-AC improves accent classification by focusing on accent embeddings.
problem Training models to predict accent type can lead to learning irrelevant features.
method GE2E-AC trains models to extract accent embeddings, making them closer for the same accent class.
result GE2E-AC outperforms baseline models trained with conventional loss.
An elementary arbitrage principle and the existence of trends in financial time series, which is based on a theorem published in 1995 by P. Cartier and Y. Perrin, lead to a new understanding of option pricing and dynamic hedging. Intricate problems related to violent behaviors of the underlying, like the existence of j…
SuTaT creates dialogue summaries for tete-a-tetes without labeled data.
problem Lack of high-quality paired dialogue-summary data.
method Unsupervised model for tete-a-tetes, modeling customer and agent roles separately.
result SuTaT outperforms on automatic and human evaluations.
This paper speeds up speech recognition by vectorizing hypotheses and speech.
problem Slowness in beam search during speech recognition.
method Vectorization of hypotheses and speech, parallelism, batch processing, shallow fusion.
result 3.7x speedup achieved by vectorizing hypotheses, 10.5x by GPU.
Generative x-vectors improve SV performance.
problem Improving text-independent speaker verification.
method Proposes a novel method to combine i-vectors and x-vectors using a transformation model derived from canonical correlation analysis.
result Generative x-vectors provide better performance than baseline systems, especially for long-duration utterances.
The Cartier-Perrin theorem, which was published in 1995 and is expressed in the language of nonstandard analysis, permits, for the first time perhaps, a clear-cut mathematical definition of the volatility of a financial asset. It yields as a byproduct a new understanding of the means of returns, of the beta coefficient…
Enhances dialogue model with persona attributes using adversarial learning.
problem Improving dialogue models to better capture speaker identity and topic.
method Adversarial learning framework with a dual discriminator system.
result phredGAN outperforms persona Seq2Seq model in various datasets.
This paper poses some basic questions about instances (hard to find) of a special problem in 3-manifold topology. "Important though the general concepts and propositions may be with the modern industrious passion for axiomatizing and generalizing has presented us...nevertheless I am convinced that the special problems …
Improves E2E ASR performance on numeric sequences with additional training data and denormalization.
problem Challenges in recognizing numeric sequences out-of-vocabulary in ASR systems.
method Uses text-to-speech for additional numeric training data and a small-footprint neural network for denormalization.
result Reduction of WER by up to a factor of 8 in the longest numeric sequences.
Recent approaches based on artificial neural networks (ANNs) have shown promising results for short-text classification. However, many short texts occur in sequences (e.g., sentences in a document or utterances in a dialog), and most existing ANN-based systems do not leverage the preceding short texts when classifying …
Speech synthesis from EEG features using RNN.
problem Speech synthesis from EEG data.
method Recurrent Neural Network (RNN) regression model to predict acoustic features from EEG features.
result Feasibility of synthesizing speech directly from EEG features demonstrated.