Multimodal disambiguation of speech recognition
View Patent ↗The present invention provides a speech recognition system combined with one or more alternate input modalities to ensure efficient and accurate text input. The speech recognition system achieves less than perfect accuracy due to limited processing power, environmental noise, and/or natural variations in speaking style. The alternate input modalities use disambiguation or recognition engines to compensate for reduced keyboards, sloppy input, and/or natural variations in writing style. The ambiguity remaining in the speech recognition process is mostly orthogonal to the ambiguity inherent in the alternate input modality, such that the combination of the two modalities resolves the recognition errors efficiently and accurately. The invention is especially well suited for mobile devices with limited space for keyboards or touch-screen input.
1. A computer-implemented method comprising:
receiving, by a mobile device, a voice input;
displaying, by the mobile device at a text insertion point of a touch screen display, a most likely interpretation of the voice input, the most likely interpretation resulting from a speech recognition process;
receiving, by the mobile device on the touch screen display, a first non-voice input that selects said most likely interpretation;
responsive to the first non-voice input, displaying for selection, by the mobile device on the touch screen display, two or more word candidates that are ordered by phonemic similarity to the most likely interpretation,
wherein the most likely interpretation and the two or more word candidates are displayed in a single window, and
wherein selection of the two or more word candidates from a list of known words is based at least in part on a confusability matrix that considers error frequency of one or more phonemes included in the most likely interpretation and positional context of the one or more phonemes within the most likely interpretation;
receiving, by the mobile device, a second non-voice input that represents a selection of an intended word candidate from among said two or more word candidates; and
automatically replacing, by the mobile device, the most likely interpretation with the intended word candidate at the text insertion point.
2. The method of claim 1 , further comprising:
for each of a series of additional non-voice inputs, determining additional word candidates that were determined from the first non-voice input.
3. The method of claim 1 , further comprising:
receiving a keypad input at a keypad of the mobile device, wherein the keypad input specifies an alphanumeric character; and
automatically interpreting the keypad input and the voice input and mutually disambiguating both inputs to produce a best interpretation of both inputs, wherein the best interpretation of both inputs is used to identify the two or more word candidates.
4. The method of claim 1 , wherein the most likely interpretation and the two or more word candidates represent alternative interpretations of the voice input.
5. The method of claim 1 , wherein the speech recognition process is performed by the mobile device.
6. The method of claim 1 , wherein the speech recognition process is performed by a server.
7. The method of claim 1 , wherein receiving the voice input is preceded by receiving activation of a push-to-talk button on the mobile device.
8. The method of claim 1 , wherein the mobile device has limited space for a keyboard or touch-screen input.
9. The method of claim 1 , further comprising:
transmitting, by the mobile device, the selection to a server, wherein said transmitting causes the server to update a database used to identify the most likely interpretation during the speech recognition process.
10. A computer program product, tangibly embodied in a non-transitory computer-readable storage medium, the computer program product including instructions operable to cause a data processing apparatus to:
receive a voice input;
display, at a text insertion point of a touch screen device, a most likely interpretation of the voice input, the most likely interpretation resulting from a speech recognition process;
receive, on the touch screen display, a first non-voice input that selects said most likely interpretation;
responsive to the first non-voice input, display for selection two or more word candidates on the touch screen display,
wherein the two or more word candidates are ordered by phonemic similarity to the most likely interpretation,
wherein the most likely interpretation and the two or more word candidates are displayed in a single window, and
wherein selection of the two or more word candidates from a list of known words is based at least in part on a confusability matrix that considers error frequency of one or more phonemes included in the most likely interpretation and positional context of the one or more phonemes within the most likely interpretation;
receive, at said non-voice input field, a second non-voice input that represents a selection of an intended word candidate from among said two or more word candidates; and
automatically replacing the most likely interpretation with the intended word candidate at the text insertion point.
11. The computer program product of claim 10 , wherein the most likely interpretation and the two or more word candidates represent alternative interpretations of the voice input.
12. The computer program product of claim 10 , wherein the speech recognition process is performed by the mobile device.
13. The computer program product of claim 10 , wherein the speech recognition process is performed by a server.
14. A mobile device including a processor configured to:
receive a voice input;
display, at a text insertion point of a touch screen display, a most likely interpretation of the voice input, the most likely interpretation resulting from a speech recognition process;
receive, on the touch screen display, a first non-voice input that selects said most likely interpretation;
responsive to the first non-voice input, display for selection two or more word candidates on the touch screen display that are ordered by phonemic similarity to the most likely interpretation,
wherein the most likely interpretation and the two or more word candidates are displayed in a single window, and
wherein selection of the two or more word candidates from a list of known words is based at least in part on a confusability matrix that considers error frequency of one or more phonemes included in the most likely interpretation and positional context of the one or more phonemes within the most likely interpretation;
receive, at said non-voice input field, a second non-voice input that represents a selection of an intended word candidate from among said two or more word candidates; and
automatically replacing the most likely interpretation with the intended word candidate at the text insertion point.
15. The mobile device of claim 14 , wherein the most likely interpretation and the two or more word candidates represent alternative interpretations of the voice input.
16. The mobile device of claim 14 , wherein the processor is further configured to perform the speech recognition process.
17. The mobile device of claim 14 , wherein the processor is further configured to receive from a server results of the speech recognition process.
18. The mobile device of claim 14 , further comprising a push-to-talk button for activating a microphone to receive the voice input.
19. The mobile device of claim 14 , wherein the touch screen display has limited space for a keyboard or touch-screen input.
20. The mobile device of claim 14 , wherein the processor is further configured to:
transmit the selection to a server, wherein said transmitting causes the server to update a database used to identify the most likely interpretation during the speech recognition process.
21. The mobile device of claim 14 , wherein considering the positional context of the one or more phonemes within the most likely interpretation includes determining whether each of the one or more phonemes is placed at the beginning, middle, or end of the most likely interpretation and whether such placement affects confusability.