Training dataset generation for speech-to-text service
AssigneeSap SE
InventorPablo Roisman
Life of this patent
Estimating the term…
Abstract
Training data for a speech-to-text service can be generated according to a variety of techniques. For example, synthetic speech audio recordings for training a speech-to-text service can be generated in an automated system via linguistic expression templates that are input to a text-to-speech service. Pre-generation characteristics and post-generation adjustments can be made. The resulting adjusted synthetic speech audio recordings can then be used for training and validation. A large number of recordings can easily be generated for development, leading to a more robust service. Domain-specific vocabulary can be supported, resulting in a trained speech-to-text service that functions well within the targeted domain.
Claims20
A computer-implemented method of automated speech-to-text training data generation comprising:
- based on a stored linguistic expression generation template following a syntax, generating a plurality of generated textual linguistic expressions, the linguistic expression generation template comprising (1) a first set of two or more alternative tokens, wherein an alternative token of the set is included within a given linguistic expression of the plurality of generated linguistic expressions;
- or (2) a variable configured to be replaced by a retrieved value of the variable in generating a generation linguistic expression of the plurality of generated linguistic expressions, wherein the generating in (1) or (2) comprises generating respective generated linguistic expressions for multiple tokens of the first set of two or more alternative tokens, or generating respective generated linguistic expressions using different values retrieved values of the variable;
- from the plurality of generated textual linguistic expressions, with a text-to-speech service, generating a plurality of synthetic speech audio recordings for developing a speech-to-text service;
- and
- training the speech-to-text service with training synthetic speech audio recordings of the plurality of generated synthetic speech audio recordings.
A computing system comprising:
- one or more processors;
- memory storing a linguistic expression generation template following a syntax;
- wherein the memory is configured to cause the one or more processors to perform operations comprising:
- based on the stored linguistic expression generation template, generating a plurality of generated textual linguistic expressions, the linguistic expression generation template comprising (1) a first set of two or more alternative tokens, wherein an alternative token of the set is included within a given linguistic expression of the plurality of generated linguistic expressions;
- or (2) a variable configured to be replaced by a retrieved value of the variable in generating a generation linguistic expression of the plurality of generated linguistic expressions, wherein the generating in (1) or (2) comprises generating respective generated linguistic expressions for multiple tokens of the first set of two or more alternative tokens or generating respective generated linguistic expressions using different values retrieved values of the variable;
- from the plurality of generated textual linguistic expressions, with a text-to-speech service, generating a plurality of synthetic speech audio recordings for developing a speech-to-text service;
- and
- training the speech-to-text service with training synthetic speech audio recordings of the plurality of generated synthetic speech audio recordings.
One or more non-transitory computer-readable media comprising computer-executable instructions that, when executed, cause a computing system to perform operations comprising:
- based on a stored linguistic expression generation template following a syntax, generating a plurality of generated textual linguistic expressions, the linguistic expression generation template comprising (1) a first set of two or more alternative tokens, wherein an alternative token of the set is included within a given linguistic expression of the plurality of generated linguistic expressions;
- or (2) a variable configured to be replaced by a retrieved value of the variable in generating a generation linguistic expression of the plurality of generated linguistic expressions, wherein the generating in (1) or (2) comprises generating respective generated linguistic expressions for multiple tokens of the first set of two or more alternative tokens or generating respective generated linguistic expressions using different values retrieved values of the variable;
- from the plurality of generated textual linguistic expressions, with a text-to-speech service, generating a plurality of synthetic speech audio recordings for developing a speech-to-text service, wherein the generating comprises adjusting a speech accent in the text-to-speech service;
- applying background noise to at least one of the plurality of synthetic speech audio recordings;
- and
- training the speech-to-text service with selected training synthetic speech audio recordings of the plurality of generated synthetic speech audio recordings.
Dependencies are inferred from the claim text.