Method for Substantial Ongoing Cumulative Voice Recognition Error Reduction

• US 20160275942A1
• Filed: 01/25/2016
• Published: 09/22/2016
• Est. Priority Date: 01/26/2015
• Status: Abandoned Application

• Current Assignee: William Drewes
• Original Assignee: William Drewes
• Inventors: Drewes, William
• Application Number:
  US15/006,092
• Publication Number:
  US 20160275942A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/0638 Interactive procedures

Abstract

In an embodiment, speech is recorded and converted to digital text based on a shared vocabulary dictionary. During the session, voice recognition errors, that is, speech that could not be automatically identified (if any exists), are identified by the system and associated with digital text. When text for the voice recognition error is identified (e.g., by an editor), the shared vocabulary dictionary is updated (so that that particular voice recognition error will not occur again), thereby improving the performance of the system for all users that use the shared vocabulary dictionary. The identification of voice recognitions errors and the updated of the vocabulary dictionary are performed on an ongoing basis, so that the performance of the system for all users continually improves.

Claims

1. A system comprising:

a processor system including one or more processors;

a storage system including one or more storage units storing at least a database and a shared vocabulary dictionary;

the storage system storing one or more machine instructions, which when implemented cause the system to implement a method including at least

recording, at the system, a user pronouncing one or more words;

automatically, by the system, while recording, recognizing and extracting one or more words on the recording, and associating the one or more words recognized with digital text, based on a shared vocabulary dictionary, the shared vocabulary dictionary being shared by multiple unrelated users;

automatically, by the system, identifying at least one voice recognition error, the voice recognition error being a sound that the system was unable to automatically associate with digital text of a word;

automatically, by the system, storing the one or more words that were recognized in the database by at least storing in the database the recording of the pronouncing of the one or more words and the digital text of the one or more words in association with each other;

automatically, by the system, storing the at least one voice recognition error by storing a sound that the system was unable to recognize and an indication that a voice recognition error occurred in association with each other;

receiving, at the system, and storing in the database said digital text to associate with the sound that the system was unable to recognize; and

updating a shared vocabulary dictionary by at least storing in the shared vocabulary dictionary the sound that the system was unable to identify and the digital text received in association with each other, so that the sound that the system was unable to recognize is recognizable by the system after the updating;

wherein the system extracts digital text of words and user pronunciations of the words extracted and updates the shared vocabulary dictionary on an ongoing basis.

2. The system of claim 1, the one or more machine instructions further include at least one machine instruction, which when implemented causes the method implemented by the system to further comprise:

prior to the recording, requesting, by the system, the user to enter identifying information; and

in response,

receiving at the system, one or more entries for the identifying information, and performing the recording.

3. The system of claim 2, the identifying information including at least a user-id of an individual speaking during a session during the recording;

of the user pronouncing the one or more words, the method implemented initiating the extracting based on the user-id.

4. The system of claim 1, wherein the automatically recognizing and associating of the one or more words recognized with the digital text includes identifying, by the system, a segment of the recording as a candidate for a word;

in response to the identifying, comparing, by the system, the one or more words to recordings of words in the shared vocabulary dictionary; and

determining, by the system, whether the segment of the recording matches one of the recordings of the words in the shared vocabulary dictionary.

5. The system of claim 4, performing an iterative process that includes at least enhancing, by the system, the segment of the recording and comparing the segment of the recording that was enhanced to the shared vocabulary dictionary,

if a match is found, exiting, by the system, the iterative process and proceeding to a next step in the method,

if a match is not found, checking one or more exit criteria, if the exit criteria is not met, repeating, by the system, the enhancing therein further enhancing the segment, and repeating, by the system, the comparing, but upon repetition, the comparing being performed with the segment after the segment was further enhanced;

if the one or more exit criteria are met, exiting, by the system, the iterative process and identifying, by the system, the segment as a voice recognition error.

6. The system of claim 1, the method further comprising:

prior to the associating of the one or more words recognized with the digital text, based on the shared vocabulary dictionary,

determining, by the system, whether an association of the one or more words recognized can be made with the digital text, based on a personal vocabulary dictionary;

if an association of the one or more words recognized can be made with the digital text, based on the personal vocabulary dictionary, associating, by the system, the one or more words recognized with the digital text, based on the personal vocabulary dictionary.

7. The system of claim 1, the shared vocabulary dictionary including at least recordings of combinations of consecutive words that are associated with a particular word as different pronunciations of the particular word, the particular word being between two other words in the recordings of the combinations of consecutive words, therein capturing variations of pronunciations of a single word according to a word that proceeds the word that is between two other words and a word that follows the word that is between two other words.

8. The system of claim 7, the method further comprising:

updating, by the system, a shared library dictionary by at least storing in the shared library dictionary the sound that the system was unable to identify and the digital text received in association with each other, so that the sound is recognizable by the system, when using the shared library dictionary for recognizing words, the shared library dictionary including recordings of individual words in association with digital text of the individual words.

9. The system of claim 1, the method further comprising:

updating, by the system, a personal vocabulary dictionary by at least storing in the personal vocabulary dictionary, in association with each other, the sound that the system was unable to identify and the digital text received, so that sound is recognizable by the system, when using the personal vocabulary dictionary for recognizing words, the personal vocabulary dictionary being a vocabulary dictionary having recordings of pronunciations of words by a single user in association with digital text of the words.

10. The system of claim 1, the shared vocabulary dictionary including at least one subject specific vocabulary dictionary.

11. The system of claim 1, the unrelated users being users that have different accounts that are not owned by a common legal entity and that are not sponsored by a common legal entity.

12. The system of claim 1, the recording including recordings from speech of multiple people, the method further comprising identifying one of the multiple people as having uttered a segment of the recording.

13. The system of claim 12, the method further comprising:

determining, by the system, that one of the multiple people is an unregistered user;

requesting, by the system, the unregistered user to read text;

recording, by the system, the text read by the unregistered user as the unregistered user reads the text;

automatically training the system to recognize words spoken by the unregistered user, based on the recording of the text read by the unregistered user.

14. The system of claim 12, the method further comprising:

determining, by the system, that at least one of the multiple people is an unregistered user and that at least one of the multiple people is a registered user;

for each registered user, prior to the associating of the one or more words recognized with the digital text, based on the shared vocabulary dictionary,

determining whether an association of the one or more words recognized can be made with the digital text, based on a personal vocabulary dictionary, and

if an association of the one or more words recognized can be made with the digital text, based on the personal vocabulary dictionary, associating the one or more words recognized with the digital text, based on the personal vocabulary dictionary; and

for each unregistered user performing the associating based on the shared vocabulary dictionary.

15. The system of claim 12, the method further comprising:

detecting, by the system, a plurality of people speaking simultaneously; and

in response, interrupting the recording, by the system, and indicating, by the system that the plurality of people speaking simultaneously was not recorded prior to continuing the recording.

16. The system of claim 1, the method further comprising:

presenting, by the system, to an editor, one or more records of speech that the system processed;

the one or more records including text of speech that was recognized as words and indications of speech that was not recognized; and

presenting, by the system, to the editor one or more tools, which when activated by the editor, cause a single indication of speech that was not recognized to be divided into multiple words that were not recognized.

17. The system of claim 16, the method further comprising:

receiving, at the system, from the editor a selection of a record;

in response to the receiving, sending to the editor information which when consumed by a viewing device causes text corresponding to speech associated with the record to be displayed, the text of the speech being interrupted with highlighted regions indicating speech that was not recognized.

18. The system of claim 16, the method further comprising providing, by the system, one or more tools for listening to the speech that was not recognized and associating text with the speech that was not recognized.

19. The system of claim 18, wherein the updating of the shared vocabulary system is based on the associating of the speech that was not recognized with text.

20. The system of claim 16, further comprising:

a set of tools including

at least one tool of the set of tools, which when activated, chooses a specific text of a word stored in the system;

at least one tool of the set of tools, which when activated causes the system to associate the speech that was not recognized with the specific text of the word that was chosen, as an alternate pronunciation of the specific text of the word that was chosen.