NAMETAG CONFUSABILITY DETERMINATION

• US 20110288867A1
• Filed: 05/18/2010
• Published: 11/24/2011
• Est. Priority Date: 05/18/2010
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors LLC (General Motors Company)
• Inventors: Chengalvarayan, Rathinavelu, Cepuran, Lawrence D.
• Granted Patent: US 8,438,028 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• CPC Class Codes: G10L 13/08 Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

A method of and system for managing nametags including receiving a command from a user to store a nametag, prompting the user to input a number to be stored in association with the nametag, receiving an input for the number from the user, prompting the user to input the nametag to be stored in association with the number, receiving an input for the nametag from the user, processing the nametag input, and calculating confusability of the nametag input in multiple individual domains including a nametag domain, a number domain, and a command domain.

Claims

1. A method of managing nametags, comprising the steps of:

(a) receiving a command from a user to store a nametag;

(b) prompting the user to input a number to be stored in association with the nametag;

(c) receiving an input for the number from the user;

(d) prompting the user to input the nametag to be stored in association with the number;

(e) receiving an input for the nametag from the user;

(f) processing the nametag input from step (e); and

(g) determining confusability of the nametag input in multiple individual domains including a nametag domain, a number domain, and a command domain.

2. The method of claim 1, wherein step (g) includes:

(g1) calculating confusability of the nametag input with previously stored nametags in the nametag domain;

(g2) prompting the user to enter a different nametag to be stored in association with the number, if the calculated confusability of step (g1) exceeds or subcedes a nametag domain confusability threshold;

otherwise, thereafter

(g3) calculating confusability of the nametag input with numbers in the number domain;

(g4) prompting the user to enter a different nametag to be stored in association with the number, if the calculated confusability of step (g3) exceeds a number domain confusability threshold;

otherwise, thereafter

(g5) calculating confusability of the nametag input with previously stored commands in the command domain;

(g6) prompting the user to utter a different nametag to be stored in association with the number, if the calculated confusability of step (g5) exceeds a command domain confusability threshold;

otherwise, thereafter

(g7) storing the nametag input in association with the nametag.

3. The method of claim 1, wherein the nametag input is an utterance from the user and the method is speaker-dependent in that step (f) includes decoding the utterance using a statistical language model (SLM) and a semantic classifier trained on the nametag domain, the number domain, and the command domain, and step (g) further includes retraining the SLM and semantic classifier using the nametag input if the nametag input was not determined to be confusable.

4. The method of claim 3, wherein step (f) includes producing an N-best list of hypotheses and associated confidence scores, and step (g) includes calculating the confusability based on at least one of the confidence scores.

5. The method of claim 1, wherein the method is speaker-independent in that the nametag input is a text entry from the user and step (f) includes converting the text entry to a sequence of phonemes, and step (g) includes calculating the confusability based on a comparison of the text entry sequence of phonemes to phonemes of entries in at least one of the domains.

6. The method of claim 5, wherein the phoneme-based confusability score is minimum if the sequence of phonemes converted from the text entry does not correspond to phonemes of any stored entry in any of the domains, and the phoneme-based confusability score is maximum if the sequence of phonemes converted from the text entry corresponds identically to a sequence of phonemes of any stored entry in any of the domains.

7. The method of claim 2, wherein the confusability thresholds are all equal to one another.

8. The method of claim 2, wherein the confusability thresholds are all different from one another.

9. A computer program product including instructions on a computer readable medium and executable by a computer processor of a system to cause the system to implement steps comprising:

(a) receiving a command from a user to store a nametag;

(b) prompting the user to input a number to be stored in association with the nametag;

(c) receiving an input for the number from the user;

(d) prompting the user to input the nametag to be stored in association with the number;

(e) receiving an input for the nametag from the user;

(f) processing the nametag input from step (e); and

(g) determining confusability of the nametag input in multiple individual domains including a nametag domain, a number domain, and a command domain.

10. The product of claim 9, wherein step (g) includes:

(g1) calculating confusability of the nametag input with previously stored nametags in the nametag domain;

(g2) prompting the user to enter a different nametag to be stored in association with the number, if the calculated confusability of step (g1) exceeds or subcedes a nametag domain confusability threshold;

otherwise, thereafter

(g3) calculating confusability of the nametag input with numbers in the number domain;

(g4) prompting the user to enter a different nametag to be stored in association with the number, if the calculated confusability of step (g3) exceeds a number domain confusability threshold;

otherwise, thereafter

(g5) calculating confusability of the nametag input with previously stored commands in the command domain;

(g6) prompting the user to utter a different nametag to be stored in association with the number, if the calculated confusability of step (g5) exceeds a command domain confusability threshold;

otherwise, thereafter

(g7) storing the nametag input in association with the nametag.

11. The product of claim 9, wherein the nametag input is an utterance from the user and the method is speaker-dependent in that step (f) includes decoding the utterance using a statistical language model (SLM) and a semantic classifier trained on the nametag domain, the number domain, and the command domain, and step (g) further includes retraining the SLM and semantic classifier using the nametag input if the nametag input is not determined to be confusable.

12. The product of claim 11, wherein step (f) includes producing an N-best list of hypotheses and associated confidence scores, and step (g) includes calculating the confusability based on at least one of the confidence scores.

13. The product of claim 9, wherein the method is speaker-independent in that the nametag input is a text entry from the user and step (f) includes converting the text entry to a sequence of phonemes, and step (g) includes calculating the confusability based on a comparison of the text entry sequence of phonemes to phonemes of entries in at least one of the domains.

14. The product of claim 13, wherein the phoneme-based confusability score is minimum if the sequence of phonemes converted from the text entry does not correspond to phonemes of any stored entry in any of the domains, and the phoneme-based confusability score is maximum if the sequence of phonemes converted from the text entry corresponds identically to a sequence of phonemes of any stored entry in any of the domains.

15. A system to manage nametags, comprising:

an interface to receive a command from a user to store a nametag, prompt the user to enter a number to be stored in association with the nametag, receive a number input for the number from the user, prompt the user to enter the nametag to be stored in association with the number, and receive a nametag input for the nametag from the user;

a processor to process the nametag input to provide a recognition result; and

a post-processor to receive the recognition result from the processor and use the recognition result as input to determine confusability of the nametag input in multiple individual domains including a nametag domain, a number domain, and a command domain.

16. The system of claim 15, wherein the nametag input is an utterance from the user and the processor decodes the utterance using a statistical language model (SLM) and a semantic classifier trained on the nametag domain, the number domain, and the command domain, and the post-processor retrains the SLM and semantic classifier using the nametag input if the nametag input is not determined to be confusable.

17. The system of claim 16, wherein the decoder produces an N-best list of hypotheses and associated confidence scores, and the post-processor calculates the confusability based on at least one of the confidence scores.

18. The system of claim 15, wherein the nametag input is a text entry from the user and the processor converts the text entry to a sequence of phonemes, and the post-processor calculates the confusability based on a comparison of the text entry sequence of phonemes to phonemes of entries in at least one of the domains.