Word correction using automatic speech recognition (ASR) incremental response
An exemplary automatic speech recognition (ASR) system may receive an audio input including a segment of speech. The segment of speech may be independently processed by general ASR and domain-specific ASR to generate multiple ASR results. A selection between the multiple ASR results may be performed based on respective confidence levels for the general ASR and domain-specific ASR. As incremental ASR is performed, a composite result may be generated based on general ASR and domain-specific ASR.
1. A method comprising:
identifying, based on a sequence aware general speech recognition application, a first plurality of candidate words corresponding to a speech-based audio input;
assigning, by the sequence aware general speech recognition application, a first respective plurality of confidence values to each of the first plurality of candidate words;
identifying, based on a sequence unaware general speech recognition application, a second plurality of candidate words corresponding to the speech-based audio input;
assigning, by the sequence unaware general speech recognition application, a second respective plurality of confidence values to each of the second plurality of candidate words;
comparing each of the first respective plurality of confidence values to each of the second respective plurality of confidence values;
selecting, based on the comparing, at least one of the first plurality of candidate words and at least one of the second plurality of candidate words; and
generating a composite plurality of words representing the speech-based audio input based on the selecting.
2. The method of claim 1 , further comprising:
identifying, from the speech-based audio input, a first speech segment and a second speech segment;
identifying, based on the sequence aware general speech recognition application, a first set of candidate words corresponding to the first segment;
assigning, by the sequence aware general speech recognition application, a first respective set of confidence values to each of the first set of candidate words;
for each of the first set of candidate words corresponding to the first segment:
identifying, based on domain-specific speech recognition, a second set of candidate words corresponding to the first segment;
assigning, by the domain-specific speech recognition, a second respective set of confidence values;
comparing each of the first respective set of confidence values to the second set of confidence values;
selecting, based on the comparing, at least one of the first set of candidate words or at least one of the second set of candidate words; and
updating a first segment of the composite plurality of words with the selected at least one of the first set of candidate words or the selected at least one of the second set of candidate words.
3. The method of claim 2 , wherein the selecting, based on the comparing, comprises selecting at least one of the first set of candidate words and at least one of the second set of candidate words.
4. The method of claim 2 , further comprising:
identifying, based on the sequence aware general speech recognition application, a third set of candidate words corresponding to the second segment;
assigning, by the sequence aware general speech recognition application, a third respective set of confidence values to each of the third set of candidate words;
for each of the third set of candidate words corresponding to the second segment:
identifying, based on domain-specific speech recognition, a fourth set of candidate words corresponding to the second segment;
assigning, by the domain-specific speech recognition, a fourth respective set of confidence values;
comparing each of the third respective set of confidence values to the fourth set of confidence values;
selecting, based on the comparing, at least one of the third set of candidate words or at least one of the fourth set of candidate words; and
updating a second segment of the composite plurality of words with the selected at least one of the third set of candidate words or the selected at least one of the fourth set of candidate words.
5. The method of claim 4 , wherein the selecting, based on the comparing, comprises selecting at least one of the third set of candidate words and at least one of the fourth set of candidate words.
6. The method of claim 1 , wherein each of the first respective plurality of confidence values and each of the second respective plurality of confidence values are based on a likelihood of a match of each of the first plurality of candidate words and each of the second plurality of candidate words, respectively, to the speech-based audio input.
7. The method of claim 2 , wherein the domain-specific specific speech recognition comprises a media guidance application.
8. The method of claim 7 , wherein the domain-specific speech recognition further comprises retrieving a plurality of metadata types for each of a plurality of media assets.
9. The method of claim 8 , the plurality of metadata types comprises at least one of title, genre, and character.
10. The method of claim 1 , further comprising:
for each of the second plurality of candidate words corresponding to the speech-based audio input:
identifying, based on domain-specific speech recognition, an alternative set of candidate words corresponding to the second plurality of candidate words;
assigning, by the domain-specific speech recognition, an alternative respective set of confidence values;
comparing each of the second respective plurality of confidence values to the alternative respective set of confidence values;
selecting, based on the comparing, at least one of the second plurality of candidate words or at least one of the alternative set of candidate words; and
updating the composite plurality of words with the selected at least one of the second plurality of candidate words or the selected at least one of the alternative set of candidate words.
11. A system comprising:
at least one input interface; and
control circuitry control circuitry configured to:
identify, based on a sequence aware general speech recognition application, a first plurality of candidate words corresponding to a speech-based audio input;
assign, by the sequence aware general speech recognition application, a first respective plurality of confidence values to each of the first plurality of candidate words;
identify, based on a sequence unaware general speech recognition application, a second plurality of candidate words corresponding to the speech-based audio input;
assign, by the sequence unaware general speech recognition application, a second respective plurality of confidence values to each of the second plurality of candidate words;
compare each of the first respective plurality of confidence values to each of the second respective plurality of confidence values;
select, based on the comparing, at least one of the first plurality of candidate words and at least one of the second plurality of candidate words; and
generate a composite plurality of words representing the speech-based audio input based on the selecting.
12. The system of claim 11 , wherein the control circuitry is further configured to:
identify, from the speech-based audio input, a first speech segment and a second speech segment;
identify, based on the sequence aware general speech recognition application, a first set of candidate words corresponding to the first segment;
assign, by the sequence aware general speech recognition application, a first respective set of confidence values to each of the first set of candidate words;
for each of the first set of candidate words corresponding to the first segment:
identify, based on domain-specific speech recognition, a second set of candidate words corresponding to the first segment;
assign, by the domain-specific speech recognition, a second respective set of confidence values;
compare each of the first respective set of confidence values to the second set of confidence values;
select, based on the comparing, at least one of the first set of candidate words or at least one of the second set of candidate words; and
update a first segment of the composite plurality of words with the selected at least one of the first set of candidate words or the selected at least one of the second set of candidate words.
13. The system of claim 12 , wherein the control circuitry configured to select, based on the comparing, is further configured to select at least one of the first set of candidate words and at least one of the second set of candidate words.
14. The system of claim 12 , wherein the control circuitry is further configured to:
identify, based on the sequence aware general speech recognition application, a third set of candidate words corresponding to the second segment;
assign, by the sequence aware general speech recognition application, a third respective set of confidence values to each of the third set of candidate words;
for each of the third set of candidate words corresponding to the second segment:
identify, based on domain-specific speech recognition, a fourth set of candidate words corresponding to the second segment;
assign, by the domain-specific speech recognition, a fourth respective set of confidence values;
compare each of the third respective set of confidence values to the fourth set of confidence values;
select, based on the comparing, at least one of the third set of candidate words or at least one of the fourth set of candidate words; and
update a second segment of the composite plurality of words with the selected at least one of the third set of candidate words or the selected at least one of the fourth set of candidate words.
15. The system of claim 14 , wherein the control circuitry configured to select, based on the comparing, is further configured to select at least one of the third set of candidate words and at least one of the fourth set of candidate words.
16. The system of claim 11 , wherein the control circuitry configured to assign each of the first respective plurality of confidence values and each of the second respective plurality of confidence values is configured to assign respective confidence values based on a likelihood of a match of each of the first plurality of candidate words and each of the second plurality of candidate words, respectively, to the speech-based audio input.
17. The system of claim 12 , wherein the control circuitry configured to identify, based on domain-specific speech recognition, is further configured to identify, based on domain-specific speech recognition, wherein the domain-specific speech recognition comprises a media guidance application.
18. The system of claim 17 , wherein the control circuitry configured to identify based on domain-specific speech recognition, is further configured to retrieve a plurality of metadata types for each of a plurality of media assets.
19. The system of claim 18 , wherein the control circuitry is further configured to retrieve a plurality of metadata types comprising at least one of title, genre, and character.
20. The system of claim 11 , wherein the control circuitry is further configured to:
for each of the second plurality of candidate words corresponding to the speech-based audio input:
identify, based on domain-specific speech recognition, an alternative set of candidate words corresponding to the second plurality of candidate words;
assign, by the domain-specific speech recognition, an alternative respective set of confidence values;
compare each of the second respective plurality of confidence values to the alternative respective set of confidence values;
select, based on the comparing, at least one of the second plurality of candidate words or at least one of the alternative set of candidate words; and
update the composite plurality of words with the selected at least one of the second plurality of candidate words or the selected at least one of the alternative set of candidate words.