Systems and methods for generating a dynamic list of hint words for automated speech recognition
Systems and methods are provided for determining hint words that improve the accuracy of automated speech recognition (ASR) systems. Hint words are typically determined in the context of a user issuing voice commands in connection with a voice interface system, however, a voice interface system may capture terms from overheard content and/or conversations. A system may determine a sliding window of hint words using set of qualifier rules. The system may capture audio, e.g., from a conversation or played back content, as a first input and decipher a plurality of words including a qualifying first term added to the hint words. The voice interface system may capture more audio as a second input and decipher a second plurality of words including a qualifying second term. The first term may be removed from the set of hint words, e.g., when the second term is added or after an expiration time.
1. A method of determining hint words for term recognition of audio input in automated speech recognition (ASR), the method comprising:
accessing one or more qualifier rules;
receiving a first audio input from a first audio source;
determining, using an automated speech recognition server, a first plurality of words from the first audio input;
determining a first term from the first plurality of words from the first audio input based on the one or more qualifier rules;
adding the first term to a set of hint words;
receiving, after the first audio input, a second audio input from a second audio source different from the first audio source;
transmitting, to the ASR server, the second audio input and the set of hint words;
determining, by the ASR server using the set of hint words, a second plurality of words from the second audio input, wherein the determined second plurality of words comprises the first term;
receiving, after the second audio input, a third audio input;
determining, using the ASR server, a third plurality of words from the third audio input;
determining a second term from the third plurality of words from the third audio input based on the one or more qualifier rules;
removing the first term from the set of hint words; and
adding the second term to the set of hint words.
2. The method of claim 1 , wherein the removing is performed in response to adding the second term.
3. The method of claim 1 , wherein the removing is performed in response to reaching a time limit.
4. The method of claim 1 , wherein the removing is performed in response to the set of hint words reaching a predetermined size.
5. The method of claim 1 , wherein the one or more qualifier rules comprises a rule that for a term at least one of the following is greater than a predetermined threshold: syllable count, phonetic matches, rhyming matches, and partial matches.
6. The method of claim 1 , wherein the one or more qualifier rules comprises a rule that for a term compared to a predetermined word there is a match of at least one of the following types: phonetic, rhyming, and partial.
7. The method of claim 1 , wherein the one or more qualifier rules comprises a rule that a term is determined to be a difficult term.
8. The method of claim 7 , wherein the difficult term is determined by calculating a difficulty score based on accessing a definition of the term.
9. The method of claim 1 , wherein the first audio source corresponds to an overheard content or conversation.
10. A system for determining hint words for term recognition in automated speech recognition (ASR), the method comprising:
input/output circuitry configured to receive a first audio input from a first audio source, receive, after the first audio input, a second audio from a second audio source different from the first audio source, and receive, after the second audio input, a third audio input; and
processing circuitry configured to:
access one or more qualifier rules;
determine, using an automated speech recognition server, a first plurality of words from the first audio input;
determine a first term from the first plurality of words from the first audio input based on the one or more qualifier rules;
add the first term to a set of hint words;
transmit to the ASR server the second audio input and the set of hint words;
determine, by the ASR server using the set of hint words, a second plurality of words from the second audio input, wherein the determined second plurality of words comprises the first term;
determining, using the ASR server, a third plurality of words from the third audio input;
determine a second term from the third plurality of words from the third audio input based on the one or more qualifier rules;
remove the first term from the set of hint words; and
add the second term to the set of hint words.
11. The system of claim 10 , wherein the processing circuitry is further configured to remove in response to adding the second term.
12. The system of claim 10 , wherein the processing circuitry is further configured to remove in response to reaching a time limit.
13. The system of claim 10 , wherein the processing circuitry is further configured to remove in response to the set of hint words reaching a predetermined size.
14. The system of claim 10 , wherein the one or more qualifier rules comprises a rule that for a term at least one of the following is greater than a predetermined threshold: syllable count, phonetic matches, rhyming matches, and partial matches.
15. The system of claim 10 , wherein the one or more qualifier rules comprises a rule that for a term compared to a predetermined word there is a match of at least one of the following types: phonetic, rhyming, and partial.
16. The system of claim 10 , wherein the one or more qualifier rules comprises a rule that a term is determined to be a difficult term.
17. The system of claim 16 , wherein the processing circuitry is further configured to determine the difficult term by calculating a difficulty score based on accessing a definition of the term.
18. The system of claim 10 , wherein the first source corresponds to an overheard content or conversation.
19. A method of determining hint words for automated speech recognition (ASR), the method comprising:
accessing one or more qualifier rules;
receiving a first audio input, wherein the first audio input comprises audio from overheard content or conversation;
determining, using an automated speech recognition server, a first plurality of words from the first audio input;
determining a first term from the first plurality of words from the first audio input based on the one or more qualifier rules;
adding the first term to a set of hint words;
receiving, after the first audio input, a second audio input;
determining the second audio input comprises the first term based at least in part on inputting to the ASR server the second audio input and the set of hint words;
receiving, after the second audio input, a third audio input;
determining, using the ASR server, a third plurality of words from the third audio input;
determining a second term from the third plurality of words from the third audio input based on the one or more qualifier rules;
removing the first term from the set of hint words; and
adding the second term to the set of hint words.