Method and apparatus of specifying and performing speech recognition operations
First Claim
Patent Images
1. A method of specifying a speech recognition operation comprising:
- receiving, on at least one computer, a recognition set from a user, the recognition set comprising one or more text words or phrases to be recognized;
automatically generating a plurality of alternate phonetic representations of each word or phrase in the recognition set;
displaying the phonetic representations to the user in a graphical user interface;
generating a plurality of speech recognition parameters for the recognition set based on said phonetic representations;
calculating, on at least one computer, an estimate of the resources used by a target system to recognize the words or phrases in the recognition set using the speech recognition parameters;
displaying the estimate to the user in the graphical user interface;
interactively modifying the phonetic representations, and in accordance therewith, modifying the speech recognition parameters, wherein the resources used by the target system are modified in accordance with the interactive modification of the phonetic representations; and
redisplaying the estimate as the phonetic representations are modified.
1 Assignment
0 Petitions
Accused Products
Abstract
A speech recognition technique is described that has the dual benefits of not requiring collection of recordings for training while using computational resources that are cost-compatible with consumer electronic products. Methods are described for improving the recognition accuracy of a recognizer by developer interaction with a design tool that iterates the recognition data during development of a recognition set of utterances and that allows controlling and minimizing the computational resources required to implement the recognizer in hardware.
747 Citations
31 Claims
-
1. A method of specifying a speech recognition operation comprising:
-
receiving, on at least one computer, a recognition set from a user, the recognition set comprising one or more text words or phrases to be recognized; automatically generating a plurality of alternate phonetic representations of each word or phrase in the recognition set; displaying the phonetic representations to the user in a graphical user interface; generating a plurality of speech recognition parameters for the recognition set based on said phonetic representations; calculating, on at least one computer, an estimate of the resources used by a target system to recognize the words or phrases in the recognition set using the speech recognition parameters; displaying the estimate to the user in the graphical user interface; interactively modifying the phonetic representations, and in accordance therewith, modifying the speech recognition parameters, wherein the resources used by the target system are modified in accordance with the interactive modification of the phonetic representations; and redisplaying the estimate as the phonetic representations are modified. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. A method of making a speech recognition device comprising:
-
receiving, on at least one computer, a recognition set from a user, the recognition set comprising one or more text words or phrases to be recognized; automatically generating a plurality of alternate phonetic representations of each word or phrase in the recognition set; displaying the phonetic representations to the user in a graphical user interface; generating a plurality of speech recognition parameters for the recognition set based on said phonetic representations; calculating, on at least one computer, an estimate of the resources used by said speech recognition device to recognize the words or phrases in the recognition set using the speech recognition parameters; displaying the estimate to the user in the graphical user interface; interactively modifying the phonetic representations, and in accordance therewith, modifying the speech recognition parameters, wherein the resources used by the speech recognition device are modified in accordance with the interactive modification of the symbolic representations; redisplaying the estimate as the phonetic representations are modified; and storing the speech recognition parameters in a memory of the speech recognition device. - View Dependent Claims (12, 13, 14, 15, 16, 17)
-
-
18. A computer-readable storage medium including software for performing a method, the method comprising:
-
receiving a recognition set from a user, the recognition set comprising one or more text words or phrases to be recognized; automatically generating a plurality of alternate phonetic representations of each word or phrase in the recognition set; displaying the phonetic representations to the user in a graphical user interface; generating a plurality of speech recognition parameters for the recognition set based on said phonetic representations; calculating an estimate of the resources used by a speech recognition device to recognize the words or phrases in the recognition set using the speech recognition parameters; displaying the estimate to the user in the graphical user interface; interactively modifying the phonetic representations, and in accordance therewith, modifying the speech recognition parameters, wherein the resources used by the speech recognition device are modified in accordance with the interactive modification of the symbolic representations; and redisplaying the estimate as the phonetic representations are modified. - View Dependent Claims (19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31)
-
Specification