FILTERING TRANSCRIPTIONS OF UTTERANCES USING RECEIVED INFORMATION TO CORRECT TRANSCRIPTION ERRORS

• US 20150255067A1
• Filed: 05/22/2015
• Published: 09/10/2015
• Est. Priority Date: 04/05/2006
• Status: Abandoned Application

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Canyon IP Holding LLC
• Inventors: White, Marc, Strohofer, Cliff
• Application Number:
  US14/720,649
• Publication Number:
  US 20150255067A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Various embodiments include methods, systems, and non-transitory computer-readable media for generating a transcription of audio data using automatic speech recognition and applying a transcription filter to a character string in the transcription (e.g., a transcribed character string) to replace the transcribed character string with another character string (e.g., a replacement character string). In some embodiments, the transcription filter applied to the transcribed character string may be ordered first in a plurality of transcription filters. In some embodiments, the transcription filter may include a finite grammar that is applied to the transcribed character string to generate the replacement character string. In some embodiments, the transcription filter may include a plurality of known character strings, and the replacement character string may be a known character string that is determined to be associated with the transcribed character filter.

Claims

1. A computer-implemented method comprising:

under control of a computing device executing specific computer-executable instructions,

obtaining audio data from a first mobile communication device;

processing the audio data using an automatic speech recognition engine to generate a transcription comprising a transcribed character string;

determining an order of a plurality of transcription filters;

applying a transcription filter ordered first in the order to the transcribed character string to generate a replacement character string; and

replacing the transcribed character string in the transcription with the replacement character string to generate a filtered transcription.

2. The computer-implemented method of claim 1, wherein:

the transcription filter ordered first in the order comprises a finite grammar;

the transcribed character string comprises a first transcribed character string and a second transcribed character string; and

applying a transcription filter ordered first in the order to the transcribed character string to generate the replacement character string comprises;

determining that, together, the first transcribed character string and the second transcribed character string comprise a sequential word grouping; and

applying the finite grammar to the sequential word grouping to generate the replacement character string.

5. The computer-implemented method of claim 1, wherein:

the transcription filter ordered first in the order comprises a plurality of known character strings;

applying the transcription filter ordered first in the order to the transcribed character string to generate a replacement character string comprises;

identifying a first known character string in the plurality of known character strings that correlates to the transcribed character string; and

determining a second known character string in the plurality of known character strings that is associated with the first known character string; and

replacing the transcribed character string in the transcription with the replacement character string comprises replacing the transcribed character string in the transcription with the second known character string.

6. The computer-implemented method of claim 5, wherein:

the plurality of known character strings comprises a plurality of obscene phrases and a plurality of censored phrases;

identifying a first known character string comprises identifying an obscene phrase in the plurality of obscene phrases that correlates to the transcribed character string;

determining a second known character string comprises determining a censored phrase in the plurality of censored phrases that is associated with the identified obscene phrase; and

replacing the transcribed character string in the transcription with the replacement character string comprises replacing the transcribed character string in the transcription with the censored phrase.

7. The computer-implemented method of claim 1, wherein:

the plurality of known character strings comprises a plurality of rhyming phrases and a plurality of greeting phrases;

identifying a first known character string comprises identifying a rhyming phrase in the plurality of rhyming phrases that rhymes with the transcribed character string;

determining a second known character string comprises determining a greeting phrase in the plurality of greeting phrases that is associated with the identified rhyming phrase; and

replacing the transcribed character string in the transcription with the replacement character string comprises replacing the transcribed character string in the transcription with the greeting phrase.

8. The computer-implemented method of claim 1, wherein:

the plurality of known character strings comprises a plurality of rhyming phrases and a plurality of closing phrases;

identifying a first known character string comprises identifying a rhyming phrase in the plurality of rhyming phrases that rhymes with the transcribed character string;

determining a second known character string comprises determining a closing phrase in the plurality of closing phrases that is associated with the identified rhyming phrase; and

replacing the transcribed character string in the transcription with the replacement character string comprises replacing the transcribed character string in the transcription with the closing phrase.

9. The computer-implemented method of claim 1, further comprising:

transmitting the filtered transcription to the first mobile communication device;

receiving a verification message from the first mobile communication device; and

transmitting the filtered transcription to a second mobile communication device in response to receiving the verification message from the first mobile communication device.

10. The computer-implemented method of claim 1, further comprising transmitting the transcription and the filtered transcription to the first mobile communication device.

11. The computer-implemented method of claim 1, wherein the transcription filter comprises at least one of a finite grammar or a plurality of known character strings.

12. The computer-implemented method of claim 1, wherein:

the filtered transcription comprises another transcribed character string; and

the computer-implemented method further comprises;

applying a transcription filter ordered second in the order to the another transcribed character string to generate another replacement character string; and

replacing the another transcribed character string in the filtered transcription with the another replacement character string.

13. The computer-implemented method of claim 1, wherein:

the plurality of transcription filters comprises a finite grammar filter and a software filter; and

the finite grammar filter is ordered before the software filter in the order of the plurality of transcription filters.

14. The computer-implemented method of claim 1, wherein:

the computer-implemented method further comprises receiving order information from the first mobile communication device; and

determining the order of the plurality of transcription filters comprises determining the order of the plurality of transcription filters from the received order information.

3. The computer-implemented method of claim 3, wherein:

the finite grammar comprises a finite grammar of currency;

the first transcribed character string represents an amount of currency; and

the second transcribed character string represents a type of currency.

4. The computer-implemented method of claim 3, wherein:

the finite grammar comprises a finite grammar of time information;

the first transcribed character string represents a numerical measurement of time; and

the second transcribed character string represents an indication of a time of day.

15. A system comprising:

a data store configured to store specific processor-executable instructions; and

a computing device in communication with the data store, the computing device comprising one or more processors configured to execute the processor-executable instructions to perform operations comprising;

obtaining audio data from a mobile communication device;

processing the audio data using an automatic speech recognition engine to generate a transcription comprising a transcribed character string;

determining an order of a plurality of transcription filters;

applying a transcription filter ordered first in the order to the transcribed character string to generate a replacement character string; and

replacing the transcribed character string in the transcription with the replacement character string to generate a filtered transcription.

16. The system of claim 15, wherein:

the transcription filter ordered first in the order comprises a finite grammar;

the transcribed character string comprises a first transcribed character string and a second transcribed character string; and

applying the transcription filter ordered first in the order to the transcribed character string to generate the replacement character string comprises;

determining that, together, the first transcribed character string and the second transcribed character string comprise a sequential word grouping; and

applying the finite grammar to the sequential word grouping to generate the replacement character string.

17. The system of claim 15, wherein:

the transcription filter ordered first in the order comprises a plurality of known character strings;

applying the transcription filter ordered first in the order to the transcribed character string to generate a replacement character string comprises;

identifying a first known character string in the plurality of known character strings that correlates to the transcribed character string; and

determining a second known character string in the plurality of known character strings that is associated with the first known character string; and

replacing the transcribed character string in the transcription with the replacement character string comprises replacing the transcribed character string in the transcription with the second known character string.

18. A non-transitory, computer-readable storage medium having stored thereon computer-executable software instructions configured to cause a processor of a computing device to perform operations comprising:

obtaining audio data from a mobile communication device;

processing the audio data using an automatic speech recognition engine to generate a transcription comprising a transcribed character string;

determining an order of a plurality of transcription filters;

applying a transcription filter ordered first in the order to the transcribed character string to generate a replacement character string; and

replacing the transcribed character string in the transcription with the replacement character string to generate a filtered transcription.

19. The non-transitory, computer-readable storage medium of claim 18, wherein:

the filtered transcription comprises another transcribed character string; and

the operations further comprise;

applying a transcription filter ordered second in the order to the another transcribed character string to generate another replacement character string; and

replacing the another transcribed character string in the filtered transcription with the another replacement character string.

20. The computer-implemented method of claim 18, wherein the operations further comprise receiving order information from the mobile communication device;

and wherein determining the order of the plurality of transcription filters comprises determining the order of the plurality of transcription filters from the order information that is received.