System and method for detecting phonetically similar imposter phrases

• US 10,269,352 B2
• Filed: 12/23/2016
• Issued: 04/23/2019
• Est. Priority Date: 12/23/2016
• Status: Active Grant

• Current Assignee: Nice Ltd
• Original Assignee: Nice Ltd
• Inventors: Morris, Robert William, Verma, Neeraj Singh
• Primary Examiner(s): Le, Thuykhanh
• Application Number:
  US15/390,026
• Publication Number:
  US 20180182378A1
• Time in Patent Office: 851 Days
• Field of Search: None
• CPC Class Codes: G06F 16/3343 using phonetics
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A system and method for detecting phonetically similar imposter phrases may include using automatic speech recognition (ASR) to search for a first phrase in a set of objects; producing a list of references by searching for the first phrase in the set of objects using phonetic search; using output produced by the ASR to determine whether or not a reference in the list points to a phrase that is the same as the first phrase; and if it is determined that the reference points to a second phrase that is different from the first phrase then marking the second phrase as a potential cause for a phrase search false positive.

Claims

1. A method identifying potential cause for a phrase search false positive when searching audio interaction objects, the method comprising:

using a processor conducting automatic speech recognition (ASR) to search for a first phrase in a set of objects to produce ASR results comprising identification of one or more objects identified by ASR as including the first phrase;

producing using the processor a list of phonetic references comprising references to one or more objects identified by phonetic search as including the first phrase by searching for the first phrase in the set of objects using phonetic search;

at the processor using output produced by the ASR to determine, using a phrase identification and query generation unit, whether or not a reference in the list of phonetic references points to a phrase that is the same as the first phrase;

if it is determined that the reference points to a second phrase that is different from the first phrase then marking by the processor the second phrase as a potential cause for the phrase search false positive;

displaying a list of potential causes for the phrase search false positive with respect to the first phrase to a user;

accepting an input from the user, the input comprising a potential cause for the phrase search false positive from the display; and

based on the input creating an operator designed to cause a query for searching the first phrase to omit the potential cause for the phrase search false positive in the input from a response of the query.

2. The method of claim 1 wherein the list of phonetic references includes, for each phrase, an identification of an object that includes the phrase, an offset of the phrase in the object and a time length of the phrase, and wherein marking the second phrase as a potential cause for a phrase search false positive includes at least one of:

verifying a reference to the phrase in the list of phonetic references and a reference to a phrase identified by the ASR reference to the same object, and

verifying, based on an offset and a time length in the list of phonetic references, that an overlap of a phrase referenced in the list of phonetic references with a phrase identified by the ASR is above a threshold.

3. The method of claim 1 wherein the list of phonetic references includes, for each referenced phrase, a confidence score value and wherein the second phrase is marked as a potential cause for a phrase search false positive if the confidence score value of the second phrase is above a threshold.

4. The method of claim 3 comprising:

if the confidence score value of the second phrase is above the threshold then including the second phrase in a set of confusable phrases.

5. The method of claim 4 wherein:

if the number of references in the list of phonetic references is below a predefined value then setting the value of the threshold to a predefined threshold value; and

if the number of references in the list of phonetic references is above the predefined value then setting the value of the threshold based on the lowest confidence score value.

6. The method of claim 4 comprising removing a phrase from the set of confusable phrases based on a match level of the phrase with the first phrase.

7. The method of claim 4 comprising removing a phrase from the set of confusable phrases based at least one of:

the number of references to the phrase in the list of phonetic references and a presence of stop words in the phrase.

8. A method identifying potential cause for a phrase search false positive when searching audio interaction objects, the method comprising:

producing using a processor automatic speech recognition (ASR) results by searching for a first phrase in a set of objects using ASR, the ASR results comprising identification of one or more objects identified by ASR as including the first phrase;

producing using the processor phonetic search results by searching for the first phrase in the set of objects using phonetic search, the phonetic search results comprising references to one or more objects identified as including the first phrase;

if a mismatch between a phrase included in the phonetic search results and a phrase included in the ASR results is found then marking the phrase included in the phonetic search as the potential cause for a phrase search false positive;

displaying a list of potential causes for the phrase search false positive with respect to the first phrase to a user;

accepting an input from the user, the input comprising a potential cause for the phrase search false positive from the display; and

based on the input creating an operator designed to cause a query for searching the first phrase to omit the potential cause for the phrase search false positive in the input from a response of the query.

9. The method of claim 8 comprising generating a search query based on the potential cause for a phrase search false positive.

10. The method of claim 8 comprising marking the phrase as a potential cause for a phrase search false positive if a confidence score value of the phrase included in the phonetic search results is above a threshold.

11. The method of claim 10 comprising:

if the confidence score value of the second phrase is above the threshold then including the second phrase in a set of confusable phrases.

12. A system comprising:

a memory; and

a controller to;

use automatic speech recognition (ASR) to search for a first phrase in a set of objects to produce ASR results comprising identification of one or more objects identified by ASR as including the first phrase;

produce a list of references by searching for the first phrase in the set of objects using phonetic search;

use output produced by the ASR to determine whether or not a reference in the list of references points to a phrase that is the same as the first phrase,

if it is determined that the reference points to a second phrase that is different from the first phrase then marking the second phrase as a potential cause for a phrase search false positive;

displaying a list of potential causes for the phrase search false positive with respect to the first phrase to a user;

accepting an input from the user, the input comprising a potential cause for the phrase search false positive from the display; and

based on the input creating an operator designed to cause a query for searching the first phrase to omit the potential cause for the phrase search false positive in the input from a response of the query.

13. The system of claim 12 wherein the list of references includes, for each phrase, an identification of an object that includes the phrase, an offset of the phrase in the object and a time length of the phrase, and wherein marking the second phrase as the potential cause for a phrase search false positive includes at least one of:

verifying a reference to the phrase in the list of references and a reference to a phrase identified by the ASR are related to the same object, and

verifying, based on an offset and a time length in the list of references, that an overlap of a phrase referenced in the list of references with a phrase identified by the ASR is above a threshold.

14. The system of claim 12 wherein the list of references includes, for each referenced phrase, a confidence score value and wherein the second phrase is marked as a potential cause for a phrase search false positive if the confidence score value for the second phrase is above a threshold.

15. The system of claim 14 wherein the controller is configured to:

if the confidence score value of the second phrase is above the threshold then include the second phrase in a set of confusable phrases.

16. The system of claim 15 wherein the controller is configured to:

if the number of references in the list of references is below a predefined value then setting the value of the threshold to a predefined threshold value; and

if the number of references in the list of references is above the predefined value then setting the value of the threshold based on the lowest confidence score value.

17. The system of claim 15 wherein the controller is configured to remove a phrase from the set of confusable phrases based on a match level of the phrase with the first phrase.

18. The system of claim 15 wherein the controller is configured to remove a phrase from the set of confusable phrases based at least one of:

the number of references to the phrase in the list of references and a presence of stop words in the phrase.