One-step repair of misrecognized recognition strings
View Patent ↗A method for use in automatic speech recognition corrects erroneous recognition elements within a recognition hypothesis. A user input is recognized as a correction hypothesis which contains various recognition elements. A non-deterministic alignment is performed to align at least a portion of the correction hypothesis with an earlier recognition hypothesis which also contains various recognition elements such that the recognition elements in the aligned portion of the correction hypothesis are determined to most likely, correspond to a range of recognition elements in the earlier recognition hypotheses. The recognition elements in the range of recognition elements in the earlier recognition hypothesis are replaced with the recognition elements in the aligned portion of the correction hypothesis.
1. A method for use in automatic speech recognition comprising:
recognizing a user input as a correction hypothesis containing a plurality of recognition elements;
performing a non-deterministic alignment of at least a portion of the correction hypothesis with an earlier recognition hypothesis containing a plurality of recognition elements such that the recognition elements in the aligned portion of the correction hypothesis are determined to most likely correspond to a range of recognition elements in the earlier recognition hypothesis;
replacing the elements in the range of recognition elements in the earlier recognition hypothesis with the recognition elements in the aligned portion of the correction hypothesis.
2. A method according to claim 1 , wherein the non-deterministic alignment favors alignments corresponding to recognition elements in the earlier recognition hypothesis that have a higher probability of being misrecognized.
3. A method according to claim 1 , wherein the non-deterministic alignment uses a Viterbi-based search.
4. A method according to claim 1 , wherein the recognition elements are organized into arrays such that the non-deterministic alignment is based on an alignment of the array elements.
5. A method according to claim 4 , wherein each recognition element is a separate array entry.
6. A method according to claim 4 , wherein the non-deterministic alignment includes using splicing markers to establish the boundaries of the aligned portion of the correction hypothesis.
7. A method according to claim 4 , wherein the non-deterministic alignment includes using splicing markers to establish boundaries of the range of recognition elements in the earlier recognition hypothesis.
8. A method according to claim 1 , wherein the earlier recognition hypothesis is an intermediate recognition hypothesis in a recognition process.
9. A method according to claim 1 , wherein the earlier recognition hypothesis is an entry on N-Best list of recognition hypotheses.
10. A method according to claim 1 , wherein the recognition elements are digits.
11. A method according to claim 1 , wherein the recognition elements are alphanumeric characters.
12. A speech recognition system adapted to use the method according to any of claims 1 - 11 .