Speech recognition apparatus & method having dynamic reference pattern adaptation
First Claim
1. A speech recognition method comprising the steps ofstoring a plurality of reference patterns representing speech to be recognized, storing in association with each reference pattern a quality value representing the effectiveness of that pattern for recognizing past incoming speech utterances,providing correction actions representing the accuracy of a speech recognition during the recognition of unknown incoming speech utterances and after training of the system for improving the accuracy of the speech recognition, anddynamically updating the quality values, during speech recognition, for at least a portion of those reference patterns used during the speech recognition process in response to said correction actions characterizing the usefulness of these reference patterns.
10 Assignments
0 Petitions
Accused Products
Abstract
A speech recognition apparatus having reference pattern adaptation stores a plurality of reference patterns representing speech to be recognized, each stored reference pattern having associated therewith a quality value representing the effectiveness of that pattern for recognizing an incoming speech utterance. The method and apparatus provide user correction actions representing the accuracy of a speech recognition, dynamically, during the recognition of unknown incoming speech utterances and after training of the system. The quality values are updated, during the speech recognition process, for at least a portion of those reference patterns used during the speech recognition process. Reference patterns having low quality values, indicative of either inaccurate representation of the unknown speech or non-use, can be deleted so long as the reference pattern is not needed, for example, where the reference pattern is the last instance of a known word or phrase. Various methods and apparatus are provided for determining when reference patterns can be deleted or added, to the reference memory, and when the scores or values associated with a reference pattern should be increased or decreased to represent the "goodness" of the reference pattern in recognizing speech.
390 Citations
20 Claims
-
1. A speech recognition method comprising the steps of
storing a plurality of reference patterns representing speech to be recognized, storing in association with each reference pattern a quality value representing the effectiveness of that pattern for recognizing past incoming speech utterances, providing correction actions representing the accuracy of a speech recognition during the recognition of unknown incoming speech utterances and after training of the system for improving the accuracy of the speech recognition, and dynamically updating the quality values, during speech recognition, for at least a portion of those reference patterns used during the speech recognition process in response to said correction actions characterizing the usefulness of these reference patterns.
-
10. A speech recognition method comprising the steps of
storing a plurality of reference patterns representing speech to be recognized, each reference pattern having associated therewith a quality value representing the effectiveness of that pattern for recognizing past incoming speech utterances, providing correction actions representing the accuracy of a speech recognition during the recognition of unknown incoming speech utterances and after training of the system for improving the accuracy of the speech recognition, associating with each unknown incoming speech utterance a recognition event record, containing information about said associated speech utterance, each said record containing at least an identification of a best candidate matching the utterance, and an updateable identification of alternative candidates matching the utterance, and an updateable identification of the method of reaching the correct answer, increasing the quality value, during speech recognition, of reference patterns associated with a correct identification of an incoming speech utterance, decreasing the quality value, during speech recognition, of reference patterns associated with an incorrect identification of an incoming utterance, adding new reference patterns to said stored reference patterns during said speech recognition process in response to said correction actions for providing additional reference patterns for recognizing unknown speech input utterances, identifying a reference pattern having a minimum quality value associated therewith, determining whether the identified reference pattern has a deletion-inhibiting status, and deleting the identified reference pattern in the absence of said deletion-inhibiting status.
-
11. A speech recognition apparatus comprising
means for storing a plurality of reference patterns representing speech to be recognized, each reference pattern having associated therewith, in said storage means, a quality value representing the effectiveness of that pattern for recognizing past incoming speech utterances, means for providing correction actions representing the accuracy of a speech recognition during the recognition of an unknown incoming speech utterance, and after training of the system, said correction actions improving the accuracy of the speech recognition, and means for dynamically updating the quality values, during speech recognition and in response to the correction actions, for at least a portion of those reference patterns used during the speech recognition process.
-
20. A speech recognition apparatus comprising
means for storing a plurality of reference patterns representing speech to be recognized, each reference pattern having associated therewith a quality value representing the effectiveness of that pattern for recognizing past incoming speech utterances, means for providing correction actions representing the accuracy of a speech recognition during the recognition of an unknown incoming speech utterance, and after training of the apparatus said correction actions improving the accuracy of the speech recognition, means for associating with each unknown incoming speech utterance input a recognition event record containing information about said associated speech utterance, each said record containing at least an identification of a best candidate matching the utterance, an identification of alternative candidates matching the utterance, and an updateable identification of the method of reaching the correct candidate, means for increasing, during speech recognition, the quality value of reference patterns associated with an incorrect identification of an incoming speech utterance, means for decreasing, during speech recognition, the quality value of reference patterns associated with an incorrect identification of an incoming utterance, means for adding new reference patterns to said stored reference patterns during said speech recognition process in response to said correction actions for providing additional reference patterns for recognizing unknown speech input utterances, means for identifying a reference pattern having a minimum quality value associated therewith, means for determining whether the identified reference pattern has a deletion-inhibiting status, and means for deleting the identified reference pattern in the absence of said deletion-inhibition status.
Specification