Simultaneous translation of open domain lectures and speeches
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.
1. 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 an end-of-sentence cue from one or more listeners, the end-of-sentence cue being commonly associated with an end of a sentence;
determining a segment boundary for a translatable segment based on the received end-of-sentence cue;
resegmenting, by the resegmentation unit, the merged partial hypotheses into the translatable segment in the first language based on the determined segment boundary; and
receiving, by a machine translation unit that is in communication with the resegmentation unit, the translatable segment in the first language from the resegmentation unit
outputting, by the machine translation unit, a translation of the spoken sound from the first speaker into a second language based on the received translatable segment.
2. The computer-implemented method of claim 1 , further comprising:
receiving, by a speech captioning unit that is in communication with the resegmentation unit, the translatable segment in the first language from the resegmentation unit; and
outputting, by the speech captioning unit, a caption of the spoken sound from the first speaker in the first language based on the received translatable segment.
3. 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;
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 an end-of-sentence cue from one or more listeners, the end-of-sentence cue being commonly associated with an end of a sentence;
determine a segment boundary for a translatable segment based on the received end-of-sentence cue; and
resegment the merged partial hypotheses into the translatable segment in the first language based on the determined segment boundary; and
a machine translation unit in communication with the resegmentation unit, wherein the machine translation unit is configured to:
receive the translatable segment in the first language from the resegmentation unit; and
output a translation of the spoken sound from the first speaker into a second language based on the received translatable segment.
4. The computer-implemented method of claim 1 , wherein the end-of-sentence cue is a pause.
5. The computer-implemented method of claim 1 , wherein the end-of-sentence cue is phraseology.
6. The computer-implemented method of claim 1 , wherein the end-of-sentence cue is a prosodic cue.
7. The computer-implemented method of claim 6 , wherein prosodic cue is based on pitch.
8. The computer-implemented method of claim 6 , wherein prosodic cue is based on loudness.
9. The computer-implemented method of claim 6 , wherein prosodic cue is based on duration.
10. The computer-implemented method of claim 1 , wherein the segment boundary for the translatable segment is determined based on an additional cue in addition to the end-of-sentence cue.
11. The computer-implemented method of claim 3 , further comprising:
receiving, by a speech index unit that is in communication with the resegmentation unit, the translatable segment in the first language from the resegmentation unit; and
outputting, by the speech index unit, an index of the spoken sound from the first speaker in the first language based on the received translatable segment.
12. The system of claim 3 , further comprising:
a speech captioning unit in communication with the resegmentation unit, wherein the speech captioning unit is configured to:
receive the translatable segment in the first language from the resegmentation unit; and
output a caption of the spoken sound from the first speaker in the first language based on the received translatable segment.
13. The system of claim 3 , further comprising:
a speech index unit in communication with the resegmentation unit, speech index unit is configured to:
receive the translatable segment in the first language from the resegmentation unit; and
output an index of the spoken sound from the first speaker in the first language based on the received translatable segment.
14. The system of claim 3 , wherein the end-of-sentence cue is a pause.
15. The system of claim 3 , wherein the end-of-sentence cue is phraseology.
16. The system of claim 3 , wherein the end-of-sentence cue is a prosodic cue.
17. The system of claim 16 , wherein prosodic cue is based on pitch.
18. The system of claim 16 , wherein prosodic cue is based on loudness.
19. The system of claim 16 , wherein prosodic cue is based on duration.
20. The system of claim 3 , wherein the segment boundary for the translatable segment is determined based on an additional cue in addition to the end-of-sentence cue.