Simultaneous Translation of Open Domain Lectures and Speeches
AssigneeFacebook, Inc.
InventorAlexander Waibel
Life of this patent
Estimating the term…
Abstract
Speech translation systems and methods for simultaneously translating speech between first and second speakers, wherein the first speaker speaks in a first language and the second speaker speaks in a second language that is different from the first language. The speech translation system may comprise a resegmentation unit that merge at least two partial hypotheses and resegments the merged partial hypotheses into a first-language translatable segment, wherein a segment boundary for the first-language translatable segment is determined based on sound from the second speaker.
Claims20
A system comprising:
- an automatic speech recognition unit configured for determining spoken sound from a first speaker in a first language and for creating a plurality of partial hypotheses of the spoken sound of the first speaker;
- and
- a resegmentation unit in communication with the automatic speech recognition unit, wherein the resegmentation unit is configured to:
- merge at least two of the partial hypotheses received from the automatic speech recognition unit;
- receive back channel information from one or more listeners, the one or more listeners different than the first speaker,
- determine a segment boundary for a translatable segment based on the received back channel information;
- and
- resegment the merged partial hypotheses into the translatable segment in the first language based on the determined segment boundary.
A computer-implemented method comprising:
- determining, by an automatic speech recognition unit, spoken sound from a first speaker in a first language;
- creating a plurality of partial hypotheses of the spoken sound of the first speaker;
- merging, by a resegmentation unit that is in communication with the automatic speech recognition unit, at least two of the partial hypotheses received from the automatic speech recognition unit;
- receiving back channel information from one or more listeners, the one or more listeners different than the first speaker,
- determining a segment boundary for a translatable segment based on the received back channel information;
- and
- resegmenting, by the resegmentation unit, the merged partial hypotheses into the translatable segment in the first language based on the determined segment boundary.
Dependencies are inferred from the claim text.