Stable output streaming speech translation system
AssigneeMicrosoft Technology Licensing, LLC
InventorsJunkun Chen, Jinyu Li, Peidong Wang, Jian Xue
Life of this patent
Estimating the term…
Abstract
A computer implemented method includes receiving speech data representative of speech in a first language The speech data is divided into chunks of speech data, each chunk comprising multiple temporally consecutive frames of acoustic information. Each temporally consecutive chunk of data is processed using beam search on each frame to identify candidate language tokens representing a second language different from the first language. A best candidate language token(s) is selected for each chunk as processed. The selected best candidate language token or tokens for each chunk of data is committed as a prefix for a next temporally consecutive chunk of data.
Claims20
A computer implemented method comprising:
- receiving speech data representative of speech in a first language;
- dividing the speech data into temporally consecutive chunks of digital speech data, each chunk comprising multiple temporally consecutive frames of acoustic information;
- processing each temporally consecutive chunk of data in a model that includes a language prediction model and encoder, wherein the encoder correlates the digital speech data to language tokens using beam search on each frame to identify candidate language tokens representing a second language different from the first language;
- selecting a best candidate language token or tokens for each chunk as processed;
- and
- committing the selected best candidate language token or tokens for each chunk of data as a prefix for a next temporally consecutive chunk of data and as a translation of the speech data into the second language.
A machine-readable storage device having instructions for execution by a processor of a machine to cause the processor to perform operations to perform a method, the operations comprising,
- receiving speech data representative of speech in a first language;
- dividing the speech data into temporally consecutive chunks of digital speech data, each chunk comprising multiple temporally consecutive frames of acoustic information;
- processing each temporally consecutive chunk of data in a model that includes a language prediction model and encoder, wherein the encoder correlates the digital speech data to language tokens using beam search on each frame to identify candidate language tokens representing a second language different from the first language;
- selecting a best candidate language token or tokens for each chunk as processed;
- and
- committing the selected best candidate language token or tokens for each chunk of data as a prefix for a next temporally consecutive chunk of data and as a translation of the speech data into the second language.
A device comprising:
- a processor;
- and
- a memory device coupled to the processor and having a program stored thereon for execution by the processor to perform operations comprising;
- receiving speech data representative of speech in a first language;
- dividing the speech data into temporally consecutive chunks of digital speech data, each chunk comprising multiple temporally consecutive frames of acoustic information;
- processing each temporally consecutive chunk of data in a model that includes a language prediction model and encoder, wherein the encoder correlates the digital speech data to language tokens using beam search on each frame to identify candidate language tokens representing a second language different from the first language;
- selecting a best candidate language token or tokens for each chunk as processed;
- and
- committing the selected best candidate language token or tokens for each chunk of data as a prefix for a next temporally consecutive chunk of data and as a translation of the speech data into the second language.
Dependencies are inferred from the claim text.