Method and apparatus for generating hint words for automated speech recognition
Systems and methods for determining hint words that improve the accuracy of automated speech recognition (ASR) systems. Hint words are determined in the context of a user issuing voice commands in connection with a voice interface system. Terms are initially taken from most frequently occurring terms in operation of a voice interface system. For example, most frequently occurring terms that arise in electronic search queries or received commands are selected. Certain of these terms are selected as hint words, and the selected hint words are then transmitted to an ASR system to assist in translation of speech to text.
1. A method of determining hint words for automated speech recognition, the method comprising:
determining, using processing circuitry, a set of logged terms from a first predetermined number of recent electronic search queries;
accessing a graph of terms;
comparing each of the set of logged terms to the graph of terms;
for each of the set of logged terms that is present in the graph of terms, selecting a second predetermined number of proximate terms from the graph of terms, so as to form a set of proximate terms corresponding to each of the set of logged terms present in the graph of terms;
selecting terms common to the sets of proximate terms, so as to form a set of hint words; and
transmitting, using the processing circuitry, the hint words to an automated speech recognition application.
2. The method of claim 1 , wherein the graph of terms is a graph of entertainment terms.
3. The method of claim 1 , wherein the selecting the second predetermined number of proximate terms further comprises, for each of the logged terms that is present in the graph of terms:
selecting nearest connected terms to the each term of the set of logged terms that is on the graph, and adding the selected nearest connected terms to the corresponding set of proximate terms; and
successively selecting next-nearest connected terms to the each term on the graph, and adding the successively selected terms to the corresponding set of proximate terms until the number of terms in the corresponding set of proximate terms is at least equal to the second predetermined number.
4. The method of claim 1 , wherein the selecting terms common to the sets of proximate terms further comprises selecting terms common to every set of proximate terms, so as to form the set of hint words.
5. The method of claim 4 , wherein the selecting terms common to the sets of proximate terms further comprises:
selecting terms common to one set of proximate terms less than every set of proximate terms, and adding the terms common to one set of proximate terms less than every set of proximate terms to the set of hint words; and
successively selecting terms common to another set of proximate terms less than every set of proximate terms, and adding the terms common to another set of proximate terms less than every set of proximate terms to the set of hint words, until the number of terms in the set of hint words is at least equal to a third predetermined number.
6. The method of claim 1 , wherein the set of logged terms comprise at least one natural language word.
7. The method of claim 1 , wherein the set of logged terms comprise at least one natural language phrase.
8. The method of claim 1 , wherein the set of hint words includes phonemes of the terms.
9. The method of claim 1 , wherein the set of logged terms comprise phonetic neighbors of the terms.
10. The method of claim 1 , wherein the first predetermined number is different from the second predetermined number.
11. A system for determining hint words for automated speech recognition, the method comprising:
a storage device; and
control circuitry configured to:
determine, using processing circuitry, a set of logged terms from a first predetermined number of recent electronic search queries;
access a graph of terms;
compare each of the set of logged terms to the graph of terms;
for each of the set of logged terms that is present in the graph of terms, select a second predetermined number of proximate terms from the graph of terms, so as to form a set of proximate terms corresponding to each of the set of logged terms present in the graph of terms;
select terms common to the sets of proximate terms, so as to form a set of hint words; and
transmit, using the processing circuitry, the hint words to an automated speech recognition application.
12. The system of claim 11 , wherein the graph of terms is a graph of entertainment terms.
13. The system of claim 11 , wherein the selecting the second predetermined number of proximate terms further comprises, for each of the logged terms that is present in the graph of terms:
selecting nearest connected terms to the each term of the set of logged terms that is on the graph, and adding the selected nearest connected terms to the corresponding set of proximate terms; and
successively selecting next-nearest connected terms to the each term on the graph, and adding the successively selected terms to the corresponding set of proximate terms until the number of terms in the corresponding set of proximate terms is at least equal to the second predetermined number.
14. The system of claim 11 , wherein the selecting terms common to the sets of proximate terms further comprises selecting terms common to every set of proximate terms, so as to form the set of hint words.
15. The system of claim 14 , wherein the selecting terms common to the sets of proximate terms further comprises:
selecting terms common to one set of proximate terms less than every set of proximate terms, and adding the terms common to one set of proximate terms less than every set of proximate terms to the set of hint words; and
successively selecting terms common to another set of proximate terms less than every set of proximate terms, and adding the terms common to another set of proximate terms less than every set of proximate terms to the set of hint words, until the number of terms in the set of hint words is at least equal to a third predetermined number.
16. The system of claim 11 , wherein the set of logged terms comprise at least one natural language word.
17. The system of claim 11 , wherein the set of logged terms comprise at least one natural language phrase.
18. The system of claim 11 , wherein the set of hint words includes phonemes of the terms.
19. The system of claim 11 , wherein the set of logged terms comprise phonetic neighbors of the terms.
20. The system of claim 11 , wherein the first predetermined number is different from the second predetermined number.