Adding words to a prefix tree for improving speech recognition
View Patent ↗An approach for improving speech recognition is provided. A processor receives a new word to add to a prefix tree. A processor determines a bonus score for a first transition from a first node to a second node in a prefix tree on condition that the first transition is included in a path of at least one transition representing the new word. A processor determines a hypothesis score for a hypothesis that corresponds to a speech sequence based on the prefix tree, where the hypothesis score adds the bonus score to an initial hypothesis score to determine the hypothesis score. In response to a determination that the hypothesis score exceeds a threshold value, a processor generates an output text sequence for the speech sequence based on the hypothesis.
1. A computer-implemented method for improving speech recognition, the computer-implemented method comprising:
receiving, by one or more processors, a new word to add to a prefix tree;
determining, by the one or more processors, a bonus score for a first transition from a first node to a second node in the prefix tree on condition that the first transition is included in a path of at least one transition representing the new word;
determining, by the one or more processors, a hypothesis score for a hypothesis that corresponds to a speech sequence based, at least in part, on the prefix tree, wherein the hypothesis score adds the bonus score to an initial hypothesis score to determine the hypothesis score; and
responsive to a determination that the hypothesis score exceeds a threshold value, generating, by the one or more processors, an output text sequence for the speech sequence based on the hypothesis.
2. The computer-implemented method of claim 1 , wherein the hypothesis score is determined by a transducer that is not trained with the new word.
3. The computer-implemented method of claim 2 , wherein the hypothesis score is further based on an adjustment on a correctness of the new word in the recognized text.
4. The computer-implemented method of claim 1 , wherein the second node is a leaf node in the prefix tree.
5. The computer-implemented method of claim 1 , further comprising:
cancelling, by the one or more processors, the bonus score in response to a determination that the hypothesis includes a second transition that is not included in the path of the at least one transition representing the new word.
6. The computer-implemented method of claim 1 , wherein the determining a bonus score further comprises:
determining, by the one or more processors, a negative bonus score for a third transition from the second node to a third node in the prefix tree on a condition that the third transition is included in the path of the at least one transition.
7. The computer-implemented method of claim 6 , further comprising:
cancelling, by the one or more processors, the bonus score of the first transition and the negative bonus score of the third transition in response to the hypothesis further determined to include a second transition that is not included in the path of the at least one transition representing the new word.
8. A computer program product for improving speech recognition, the computer program product comprising:
one or more computer-readable storage media and program instructions stored on the one or more computer-readable storage media, the program instructions comprising:
program instructions to receive a new word to add to a prefix tree;
program instructions to determine a bonus score for a first transition from a first node to a second node in the prefix tree on condition that the first transition is included in a path of at least one transition representing the new word;
program instructions to determine a hypothesis score for a hypothesis that corresponds to a speech sequence based, at least in part, on the prefix tree, wherein the hypothesis score adds the bonus score to an initial hypothesis score to determine the hypothesis score; and
responsive to a determination that the hypothesis score exceeds a threshold value, program instructions to generate an output text sequence for the speech sequence based on the hypothesis.
9. The computer program product of claim 8 , wherein the hypothesis score is determined by a transducer that is not trained with the new word.
10. The computer program product of claim 9 , wherein the hypothesis score is further based on an adjustment on a correctness of the new word in the recognized text.
11. The computer program product of claim 8 , wherein the second node is a leaf node in the prefix tree.
12. The computer program product of claim 8 , further comprising:
program instructions to cancel the bonus score in response to a determination that the hypothesis includes a second transition that is not included in the path of the at least one transition representing the new word.
13. The computer program product of claim 8 , wherein the determining a bonus score further comprises:
program instructions to determine a negative bonus score for a third transition from the second node to a third node in the prefix tree on a condition that the third transition is included in the path of the at least one transition.
14. The computer program product of claim 13 , further comprising:
program instructions to determine cancel the bonus score of the first transition and the negative bonus score of the third transition in response to the hypothesis further determined to include a second transition that is not included in the path of the at least one transition representing the new word.
15. A computer system for improving speech recognition, the computer system comprising:
one or more computer processors;
one or more computer readable storage media; and
program instructions stored on the computer readable storage media for execution by at least one of the one or more processors, the program instructions comprising:
program instructions to receive a new word to add to a prefix tree;
program instructions to determine a bonus score for a first transition from a first node to a second node in the prefix tree on condition that the first transition is included in a path of at least one transition representing the new word;
program instructions to determine a hypothesis score for a hypothesis that corresponds to a speech sequence based, at least in part, on the prefix tree, wherein the hypothesis score adds the bonus score to an initial hypothesis score to determine the hypothesis score; and
responsive to a determination that the hypothesis score exceeds a threshold value, program instructions to generate an output text sequence for the speech sequence based on the hypothesis.
16. The computer system of claim 15 , wherein the hypothesis score is determined by a transducer that is not trained with the new word.
17. The computer system of claim 16 , wherein the hypothesis score is further based on an adjustment on a correctness of the new word in the recognized text.
18. The computer system of claim 15 , wherein the second node is a leaf node in the prefix tree.
19. The computer system of claim 15 , further comprising:
program instructions to cancel the bonus score in response to a determination that the hypothesis includes a second transition that is not included in the path of the at least one transition representing the new word.
20. The computer system of claim 15 , wherein the determining a bonus score further comprises:
program instructions to determine a negative bonus score to a third transition from the second node to a third node in the prefix tree on a condition that the third transition is included in the path of the at least one transition.