Adaptive hosted text to speech processing
Techniques are provided performing text-to-speech translation in situations in which the input texts may contain unanticipated content. According to one aspect of the invention, text-to-speech services are provided by splitting a text into segments that include anticipated-content segments and unanticipated-content segments. Speech for the anticipated-content segments is generated based on pre-recorded sound recordings that correspond to the anticipated-content segments. Speech for the unanticipated-content segments is generated using speech synthesis. Usage statistics are recorded. The usage statistics identify which segments are contained in texts that are translated using the text-to-speech services. In one embodiment, the usage statistics indicate frequency of use of unanticipated-content segments and, based on the usage statistics, a set of unanticipated-content segments for which to make recordings is selected. In another embodiment, the usage statistics indicate frequency of use of anticipated-content segments, and a set of anticipated-content segments is selected based on the usage statistics. The recordings associated with the selected anticipated-content segments are then removed.
1. A method of providing text-to-speech services, the method comprising the steps of:
splitting a text into segments that include anticipated-content segments and unanticipated-content segments,
wherein each of the anticipated-content segments have previously satisfied criteria for being pre-recorded, and
wherein each of the unanticipated-content segments are not within the anticipated-content segments;
generating speech for said anticipated-content segments based on pre-recorded sound recordings that correspond to said anticipated-content segments;
generating speech for said unanticipated-content segments using speech synthesis;
monitoring usage of a particular segment of said segments by said text-to-speech services,
wherein said particular segment is one of an anticipated-content-segment and an unanticipated-content-segment; and
based on the usage of said particular segment by said text-to-speech services, recategorizing said particular segment to the other of said anticipated-content-segment and said unanticipated-content-segment.
2. The method of claim 1 comprising the steps of storing usage statistics that identify which segments are contained in texts that are translated using said text-to-speech services.
3. The method of claim 2 wherein the usage statistics indicate frequency of use of at least a set of said segments.
4. The method of claim 3 wherein:
the usage statistics indicate frequency of use of unanticipated-content segments; and
the method includes the step of selecting, based on said usage statistics, a set of unanticipated-content segments for which to make recordings.
5. The method of claim 4 wherein the step of selecting a set of unanticipated-content segments includes selecting a set of unanticipated-content segments that were most frequently used during a time period.
6. The method of claim 3 wherein:
the usage statistics indicate frequency of use of anticipated-content segments; and
the method includes the steps of
selecting a set of anticipated-content segments based on said usage statistics; and
removing recordings associated with the selected anticipated-content segments.
7. The method of claim 6 wherein the step of selecting a set of anticipated-content segments includes selecting a set of anticipated-content segments that were least frequently used during a period of time.
8. The method of claim 1 further comprising the steps of:
recording a plurality of recordings for a particular anticipated-segment;
storing data that indicates rules for selecting between said plurality of recordings; and
when said text contains said particular anticipated-content segment, applying the rules indicated in said data to select one of said plurality of recordings; and
generating speech for said particular anticipated-segment using said selected recording.
9. The method of claim 8 wherein:
the text is from a particular source; and
the step of applying the rules includes determining which of said plurality of recordings to select based at least in part on identity of said particular source.
10. The method of claim 1 wherein:
the text is from one of a plurality of text sources managed by a plurality of parties; and
the text-to-speech services are provided by a host, separate from said plurality of parties, that is connected to said text sources over a network system.
11. The method of claim 10 wherein the text sources are web pages that contain text, and said network system is the World Wide Web.
12. The method of claim 8 wherein:
the particular anticipated-content segment appears in a particular context within said text; and
the step of applying the rules includes determining which of said plurality of recordings to select based at least in part on said particular context.
13. A computer-readable medium carrying instructions for providing text-to-speech services, the instructions including instructions for performing the steps of:
splitting a text into segments that include anticipated-content segments and unanticipated-content segments,
wherein each of the anticipated-content segments have previously satisfied criteria for being pre-recorded, and
wherein each of the unanticipated-content segments are not within the anticipated-content segments;
generating speech for said anticipated-content segments based on pre-recorded sound recordings that correspond to said anticipated-content segments;
generating speech for said unanticipated-content segments using speech synthesis, monitoring usage of a particular segment of said segments by said text-to-speech services,
wherein said particular segment is one of an anticipated-content-segment and an unanticipated-content-segment; and
based on the usage of said particular segment by said text-to-speech services, recategorizing said particular segment to the other of said anticipated-content-segment and said unanticipated-content-segment.
14. The computer-readable medium of claim 13 comprising the steps of storing usage statistics that identify which segments are contained in texts that are translated using said text-to-speech services.
15. The computer-readable medium of claim 14 wherein the usage statistics indicate frequency of use of at least a set of said segments.
16. The computer-readable medium of claim 15 wherein:
the usage statistics indicate frequency of use of unanticipated-content segments; and
the computer-readable medium includes the step of selecting, based on said usage statistics, a set of unanticipated-content segments for which to make recordings.
17. The computer-readable medium of claim 16 wherein the step of selecting a set of unanticipated-content segments includes selecting a set of unanticipated-content segments that were most frequently used during a time period.
18. The computer-readable medium of claim 15 wherein:
the usage statistics indicate frequency of use of anticipated-content segments; and
the computer-readable medium includes the steps of
selecting a set of anticipated-content segments based on said usage statistics; and
removing recordings associated with the selected anticipated-content segments.
19. The computer-readable medium of claim 18 wherein the step of selecting a set of anticipated-content segments includes selecting a set of anticipated-content segments that were least frequently used during a period of time.
20. The computer-readable medium of claim 13 further comprising the steps of:
recording a plurality of recordings for a particular anticipated-segment;
storing data that indicates rules for selecting between said plurality of recordings; and
when said text contains said particular anticipated-content segment, applying the rules indicated in said data to select one of said plurality of recordings; and
generating speech for said particular anticipated-segment using said selected recording.
21. The computer-readable medium of claim 20 wherein:
the text is from a particular source; and
the step of applying the rules includes determining which of said plurality of recordings to select based at least in part on identity of said particular source.
22. The computer-readable medium of claim 20 wherein:
the particular anticipated-content segment appears in a particular context within said text; and
the step of applying the rules includes determining which of said plurality of recordings to select based at least in part on said particular context.