Methods and System for Grammar Fitness Evaluation as Speech Recognition Error Predictor

• US 20110282667A1
• Filed: 05/14/2010
• Published: 11/17/2011
• Est. Priority Date: 05/14/2010
• Status: Active Grant

• Current Assignee: Sony Interactive Entertainment Inc. (Sony Group Corp.)
• Original Assignee: Sony Computer Entertainment Incorporated (Sony Group Corp.)
• Inventors: Hernandez-Abrego, Gustavo A.
• Granted Patent: US 8,560,318 B2
• Time in Patent Office: Days
• US Class Current: 704/254
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules

Abstract

A plurality of statements are received from within a grammar structure. Each of the statements is formed by a number of word sets. A number of alignment regions across the statements are identified by aligning the statements on a word set basis. Each aligned word set represents an alignment region. A number of potential confusion zones are identified across the statements. Each potential confusion zone is defined by words from two or more of the statements at corresponding positions outside the alignment regions. For each of the identified potential confusion zones, phonetic pronunciations of the words within the potential confusion zone are analyzed to determine a measure of confusion probability between the words when audibly processed by a speech recognition system during the computing event. An identity of the potential confusion zones across the statements and their corresponding measure of confusion probability are reported to facilitate grammar structure improvement.

Claims

1. A computer implemented method for evaluating potential confusion within a grammar structure for a set of statements to be used in speech recognition during a computing event, comprising:

receiving a plurality of statements from within a grammar structure, each of the plurality of statements formed by a number of word sets;

identifying a number of alignment regions across the plurality of statements by aligning the plurality of statements on a word set basis, wherein each aligned word set represents an alignment region;

identifying a number of potential confusion zones across the plurality of statements, wherein each potential confusion zone is defined by words from two or more of the plurality of statements at corresponding positions outside the number of alignment regions;

for each of the identified potential confusion zones, analyzing phonetic pronunciations of the words within the potential confusion zone to determine a measure of confusion probability between the words when audibly processed by a speech recognition system during the computing event; and

generating a report to convey an identity of the potential confusion zones across the plurality of statements and their corresponding measure of confusion probability.

2. A computer implemented method as recited in claim 1, wherein the computer implemented method is performed without auditory input.

3. A computer implemented method as recited in claim 1, wherein each word set includes one or more words.

4. A computer implemented method as recited in claim 1, wherein identifying the number of alignment regions across the plurality of statements includes determining a best overall alignment of identical word sets across the plurality of statements.

5. A computer implemented method as recited in claim 1, wherein analyzing phonetic pronunciations of the words within the potential confusion zone includes,

performing a phoneme-level alignment across phonemes of the words of the plurality of statements within the potential confusion zone, and

computing a phonetic accuracy value for the potential confusion zone based on the phoneme-level alignment, wherein the phonetic accuracy value corresponds to the measure of confusion probability between the words of the plurality of statements within the potential confusion zone when audibly processed by the speech recognition system during the computing event.

6. A computer implemented method as recited in claim 5, wherein a phoneme is a minimal distinct unit of a sound system of a language.

7. A computer implemented method as recited in claim 6, wherein performing the phoneme-level alignment includes determining the best overall alignment of identical phonemes of the words within the potential confusion zone across the plurality of statements.

8. A computer implemented method as recited in claim 1, further comprising:

applying a confusion probability threshold value when generating the report, such that only potential confusion zones whose measure of confusion probability is greater than or equal to the confusion probability threshold value are identified in the report.

9. A computer implemented method as recited in claim 1, further comprising:

terminating a processing of the plurality of statements within the computer-implemented method when the number of alignment regions identified across the plurality of statements is zero.

10. A computer implemented method as recited in claim 1, further comprising:

terminating a processing of the plurality of statements within the computer-implemented method when the plurality of statements are identical to each other.

11. A system for evaluating potential confusion within a grammar structure for a set of statements to be used in speech recognition during a computing event, comprising:

an input module defined to receive a plurality of statements from within a grammar structure, each of the plurality of statements formed by a number of word sets;

a word-level alignment module defined to identify a number of alignment regions across the plurality of statements received by the input module, wherein an alignment region corresponds to identical word sets aligned across the plurality of statements;

a confusion zone identification module defined to identify a number of potential confusion zones across the plurality of statements, wherein each potential confusion zone is defined by words from two or more of the plurality of statements at corresponding positions outside the number of alignment regions identified by the word-level alignment module; and

a confusion probability analysis module defined to analyze phonetic pronunciations of the words within a given potential confusion zone identified by the confusion zone identification module to determine a measure of confusion probability between the words within the given potential confusion zone when audibly processed by a speech recognition system during the computing event.

12. A system as recited in claim 11, wherein each word set includes one or more words.

13. A system as recited in claim 11, wherein the word-level alignment module is defined to determine a best overall alignment of identical word sets across the plurality of statements.

14. A system as recited in claim 11, wherein the confusion probability analysis module is defined to perform a phoneme-level alignment across phonemes of the words of the plurality of statements within the given potential confusion zone, and

compute a phonetic accuracy value based on the phoneme-level alignment, wherein the phonetic accuracy value corresponds to the measure of confusion probability between the words within the given potential confusion zone when audibly processed by the speech recognition system during the computing event.

15. A system as recited in claim 14, wherein a phoneme is a minimal distinct unit of a sound system of a language.

16. A system as recited in claim 15, wherein the confusion probability analysis module is defined to determine the best overall alignment of identical phonemes of the words within the given potential confusion zone across the plurality of statements when performing the phoneme-level alignment.

17. A system as recited in claim 11, further comprising:

an output module defined to generate a report to convey an identity of the potential confusion zones across the plurality of statements and their corresponding measure of confusion probability.

18. A system as recited in claim 17, wherein the output module is defined to apply a confusion probability threshold value to generate the report, such that only potential confusion zones whose measure of confusion probability is greater than or equal to the confusion probability threshold value are identified in the report.

19. A computer readable medium including program instructions for evaluating potential confusion within a grammar structure for a set of statements to be used in speech recognition during a computing event, comprising:

program instructions for receiving a plurality of statements from within a grammar structure, each of the plurality of statements formed by a number of word sets;

program instructions for identifying a number of alignment regions across the plurality of statements by aligning the plurality of statements on a word set basis, wherein each aligned word set represents an alignment region;

program instructions for identifying a number of potential confusion zones across the plurality of statements, wherein each potential confusion zone is defined by words from two or more of the plurality of statements at corresponding positions outside the number of alignment regions; and

program instructions for analyzing phonetic pronunciations of the words within each potential confusion zone to determine a measure of confusion probability between the words within each potential confusion zone when audibly processed by a speech recognition system during the computing event.

20. A computer readable medium as recited in claim 19, further comprising:

program instructions for generating a report to convey an identity of the potential confusion zones across the plurality of statements and their corresponding measure of confusion probability.