International Business Machines Corporation

Nuance Communications, Inc.

The Ohio State University

Ohio State Innovation Foundation

Brian E D Kingsbury

George Andrei Saon

Hong-Kwang Jeff Kuo

Samuel Jonathan Thomas

Tara N Sainath

Lidia Mangu

Zoltan Tueske

Xiaodong Cui

Michael Alan Picheny

Vaibhava Goel

Kartik Audhkhasi

Hagen Soltau

Bhuvana Ramabhadran

Gakuto Kurata

Lior Horesh

Jatin Ganhotra

Eric Fosler-Lussier

Vishal Sunder

David Nahamoo

Gary Dale Carpenter

Sachindra Joshi

Kush Raj Varshney

Harry William Printz

Etienne Marcheret

David C Haws

Vikas Sindhwani

Yoanna Baumgartner

Petr Fousek

Vijayaditya Peddinti

Yinghui Huang

Richard Siegmund, Jr

Shay Maymon

Hazar Yueksel

Abdel-rahman Mohamed

One or more systems, devices, computer program products and/or computer-implemented methods of use provided herein relate to length perturbation techniques for improving generalization of DNN acoustic models. A computer-implemented system can comprise a memory that can store computer executable components. The computer-implemented system can further comprise a processor that can execute the computer executable components stored in the memory, wherein the computer executable components can comprise a frame skipping component that can remove one or more frames from an acoustic utterance via frame skipping. The computer executable components can further comprise a frame insertion component that can insert one or more replacement frames into the acoustic utterance via frame insertion to replace the one or more frames with the one or more replacement frames to enable length perturbation of the acoustic utterance.

Length perturbation techniques for improving generalization of deep neural network acoustic models

Systems, computer-implemented methods, and computer program products to facilitate fine-grained textual knowledge transfer to improve speech recognition and understanding are provided. According to an embodiment, a system can comprise a processor that executes components stored in memory. The computer executable components comprise deriving component that can derive one or more speech-based embeddings from an utterance via a speech encoder. The computer executable components can comprise a cross-attention component that can align, at a token level, one or more large language model (LLM) based sentence embeddings with the one or more speech-based embeddings. The computer executable components can comprise a loss component that can combine an alignment loss and an automatic speech recognition (ASR) loss.

Textual knowledge transfer for improved speech recognition and understanding

Features of two or more single speaker utterances are concatenated together and corresponding labels of the two or more single speaker utterances are concatenated together. Single speaker acoustic embeddings for each of the single speaker utterances of the concatenated single speaker utterances are generated using a single speaker teacher encoder network. An enhanced model is trained on the concatenated single speaker utterances using a classification loss Land a representation similarity loss L, the representation similarity loss Ldefined to influence an embedding derived from the concatenated single speaker utterances, the influence being based on the single speaker acoustic embeddings derived from the single speaker teacher encoder network.

Multi-speaker data augmentation for improved end-to-end automatic speech recognition

Mechanisms are provided for performing machine learning training of a computer model. A perturbation generator generates a modified training data comprising perturbations injected into original training data, where the perturbations cause a data corruption of the original training data. The modified training data is input into a prediction network of the computer model and processing the modified training data through the prediction network to generate a prediction output. Machine learning training is executed of the prediction network based on the prediction output and the original training data to generate a trained prediction network of a trained computer model. The trained computer model is deployed to an artificial intelligence computing system for performance of an inference operation.

Reducing exposure bias in machine learning training of sequence-to-sequence transducers

Audio signals representing a current utterance in a conversation and a dialog history including at least information associated with past utterances corresponding to the current utterance in the conversation can be received. The dialog history can be encoded into an embedding. A spoken language understanding neural network model can be trained to perform a spoken language understanding task based on input features including at least speech features associated with the received audio signals and the embedding. An encoder can also be trained to encode a given dialog history into an embedding. The spoken language understanding task can include predicting a dialog action of an utterance. The spoken language understanding task can include predicting a dialog intent or overall topic of the conversation.

Integrating dialog history into end-to-end spoken language understanding systems

Systems, computer-implemented methods, and computer program products to facilitate end to end integration of dialogue history for spoken language understanding are provided. According to an embodiment, a system can comprise a processor that executes components stored in memory. The computer executable components comprise a conversation component that encodes speech-based content of an utterance and text-based content of the utterance into a uniform representation.

End-to-end integration of dialog history for spoken language understanding

Training data can be received, which can include pairs of speech and meaning representation associated with the speech as ground truth data. The meaning representation includes at least semantic entities associated with the speech, where the spoken order of the semantic entities is unknown. The semantic entities of the meaning representation in the training data can be reordered into spoken order of the associated speech using an alignment technique. A spoken language understanding machine learning model can be trained using the pairs of speech and meaning representation having the reordered semantic entities. The meaning representation, e.g., semantic entities, in the received training data can be perturbed to create random order sequence variations of the semantic entities associated with speech. Perturbed meaning representation with associated speech can augment the training data.

Training end-to-end spoken language understanding systems with unordered entities

A method and system of training a spoken language understanding (SLU) model includes receiving natural language training data comprising (i) one or more speech recording, and (ii) a set of semantic entities and/or intents for each corresponding speech recording. For each speech recording, one or more entity labels and corresponding values, and one or more intent labels are extracted from the corresponding semantic entities and/or overall intent. A spoken language understanding (SLU) model is trained based upon the one or more entity labels and corresponding values, and one or more intent labels of the corresponding speech recordings without a need for a transcript of the corresponding speech recording.

End-to-end spoken language understanding without full transcripts

Techniques for classifier generalization in a supervised learning process using input encoding are provided. In one aspect, a method for classification generalization includes: encoding original input features from at least one input sample {right arrow over (x)}with a uniquely decodable code using an encoder E(⋅) to produce encoded input features E({right arrow over (x)}), wherein the at least one input sample {right arrow over (x)}comprises uncoded input features; feeding the uncoded input features and the encoded input features E({right arrow over (x)}) to a base model to build an encoded model; and learning a classification function {tilde over (C)}(⋅) using the encoded model, wherein the classification function {tilde over (C)}(⋅) learned using the encoded model is more general than that learned using the uncoded input features alone.

Input encoding for classifier generalization

A computer-implemented method for customizing a recurrent neural network transducer (RNN-T) is provided. The computer implemented method includes synthesizing first domain audio data from first domain text data, and feeding the synthesized first domain audio data into a trained encoder of the recurrent neural network transducer (RNN-T) having an initial condition, wherein the encoder is updated using the synthesized first domain audio data and the first domain text data. The computer implemented method further includes synthesizing second domain audio data from second domain text data, and feeding the synthesized second domain audio data into the updated encoder of the recurrent neural network transducer (RNN-T), wherein the prediction network is updated using the synthesized second domain audio data and the second domain text data. The computer implemented method further includes restoring the updated encoder to the initial condition.

Customization of recurrent neural network transducers for speech recognition

A processor-implemented method trains an automatic speech recognition system using speech data and text data. A computer device receives speech data, and generates a spectrogram based on the speech data. The computing device receives text data associated with an entire corpus of text data, and generates a textogram based upon the text data. The computing device trains an automatic speech recognition system using the spectrogram and the textogram.

Integrating text inputs for training and adapting neural network transducer ASR models

A computer-implemented method of building a multilingual acoustic model for automatic speech recognition in a low resource setting includes training a multilingual network on a set of training languages with an original transcribed training data to create a baseline multilingual acoustic model. Transliteration of transcribed training data is performed by processing through the multilingual network a plurality of multilingual data types from the set of languages, and outputting a pool of transliterated data. A filtering metric is applied to the pool of transliterated data output to select one or more portions of the transliterated data for retraining of the acoustic model. Data augmentation is performed by adding one or more selected portions of the output transliterated data back to the original transcribed training data to update training data. The training of a new multilingual acoustic model through the multilingual network is performed using the updated training data.

Transliteration based data augmentation for training multilingual ASR acoustic models in low resource settings

In an approach to soft-forgetting training, one or more computer processors train a first model utilizing one or more training batches wherein each training batch of the one or more training batches comprises one or more blocks of information. The one or more computer processors, responsive to a completion of the training of the first model, initiate a training of a second model utilizing the one or more training batches. The one or more computer processors jitter a random block size for each block of information for each of the one or more training batches for the second model. The one or more computer processors unroll the second model over one or more non-overlapping contiguous jittered blocks of information. The one or more computer processors, responsive to the unrolling of the second model, reduce overfitting for the second model by applying twin regularization.

Soft-forgetting for connectionist temporal classification based automatic speech recognition

Techniques for learning front-end speech recognition parameters as part of training a neural network classifier include obtaining an input speech signal, and applying front-end speech recognition parameters to extract features from the input speech signal. The extracted features may be fed through a neural network to obtain an output classification for the input speech signal, and an error measure may be computed for the output classification through comparison of the output classification with a known target classification. Back propagation may be applied to adjust one or more of the front-end parameters as one or more layers of the neural network, based on the error measure.

Learning front-end speech recognition parameters within neural network training

A method for training a deep neural network, comprises receiving and formatting speech data for the training, preconditioning a system of equations to be used for analyzing the speech data in connection with the training by using a non-fixed point quasi-Newton preconditioning scheme, and employing flexible Krylov subspace solvers in response to variations in the preconditioning scheme for different iterations of the training.

Growing community of inventors

Cortlandt Manor, NY, United States of America

Brian E D Kingsbury