Age-sensitive automatic speech recognition
Systems and methods are described to receive a query from a user and provide a reply that is appropriate for an age group of the user. A query for a media asset is received, where such query comprises an inputted term, and the query is determined to be received from a user belonging to a first age group. A context of the inputted term within the query is identified, and in response to the determining, based on the identified context, that the inputted term of the query is inappropriate for the first age group, a replacement term for the inputted term that is related to the inputted term and is appropriate for the first age group in the context of the query is identified. The query is modified to replace the inputted term with the identified replacement term, and a reply to the modified query is generated for output.
1. A method comprising:
training a first machine learning model to accept as input a first query from a user belonging to a first age group and a context of a term within the first query and output a first replacement term, wherein the term within the first query is inappropriate for the first age group within the context of the first query;
training a second machine learning model to accept as input the first query and the context of the term within the first query, and output a second replacement term;
receiving a query for a media asset, wherein the query comprises an inputted term;
determining that the query was received from a user belonging to the first age group;
identifying a context of the inputted term within the query;
determining, based on the identified context, whether the inputted term of the query is inappropriate for the first age group;
in response to the determining that the inputted term of the query is inappropriate for the first age group:
identifying a replacement term for the inputted term that (a) is related to the inputted term and (b) is appropriate for the first age group in the context of the query, wherein the identifying the replacement term for the inputted term comprises:
inputting the query and the context of the inputted term within the context of the query into each of the first machine learning model and the second machine learning model to output a first replacement term semantically similar to the inputted term and a second replacement term phonetically similar to the inputted term from the first machine learning model and the second machine learning model, respectively;
comparing a confidence score of the first replacement term to a confidence score of the second replacement term; and
identifying the replacement term as the first replacement term or the second replacement term based on the comparing;
modifying the query to replace the inputted term with the identified replacement term; and
generating for output a reply to the modified query.
2. The method of claim 1 , wherein the query is a voice query, the method further comprising:
transcribing the voice query to text; wherein
modifying the query comprises modifying the transcribed text of the query by replacing the inputted term with the replacement term.
3. The method of claim 1 , wherein the first replacement term output by the first machine learning model is semantically similar to the inputted term.
4. The method of claim 1 , wherein the second replacement term output by the second machine learning model is phonetically similar to the inputted term.
5. The method of claim 1 , wherein determining whether the inputted term of the query is inappropriate for the first age group further comprises:
parsing each respective term of the query and marking each respective term as either appropriate for the first age group or inappropriate for the first age group.
6. The method of claim 1 , wherein determining the inputted term of the query is inappropriate for the first age group comprises:
determining that the inputted term matches a term in a list of terms marked as inappropriate for the first age group in the identified context.
7. The method of claim 6 , wherein the list of terms marked as inappropriate for the first age group in the identified context comprises a list of commonly misused terms by users in the first age group in the identified context.
8. The method of claim 6 , wherein the list of terms marked as inappropriate for the first age group in the identified context comprises a list of commonly mispronounced terms by users in the first age group in the identified context.
9. A system comprising:
input/output circuitry configured to:
receive a query for a media asset, wherein the query comprises an inputted term; and
control circuitry configured to:
train a first machine learning model to accept as input a first query from a user belonging to the first age group and a context of a term within the first query and output a first replacement term, wherein the term within the first query is inappropriate for the first age group within the context of the first query;
train a second machine learning model to accept as input the first query and the context of the term within the first query, and output a second replacement term;
identify a context of the inputted term within the query;
determine, based on the identified context and based on whether a reply to the query would comprise a reference to content with metadata indicating that the content is inappropriate for the first age group if the query is not modified, whether the inputted term of the query is inappropriate for the first age group;
identify a replacement term for the inputted term by:
inputting the query and the context of the inputted term within the context of the query into each of the first machine learning model and the second machine learning model to output a first replacement term semantically similar to the inputted term and a second replacement term phonetically similar to the inputted term from the first machine learning model and the second machine learning model, respectively;
comparing a confidence score of the first replacement term to a confidence score of the second replacement term; and
identifying the replacement term as the first replacement term or the second replacement term based on the comparing;
modify the query to replace the inputted term with the identified replacement term; and
generate for output a reply to the modified query.
10. The system of claim 9 , wherein the query is a voice query, and the control circuitry is further configured to:
transcribe the voice query to text; wherein
modifying the query comprises modifying the transcribed text of the query by replacing the inputted term with the replacement term.
11. The system of claim 9 , wherein the first replacement term output by the first machine learning model is semantically similar to the inputted term.
12. The system of claim 9 , wherein the second replacement term output by the second machine learning model is phonetically similar to the inputted term.
13. The system of claim 9 , wherein the control circuitry is configured to determine whether the inputted term of the query is inappropriate for the first age group by:
parsing each respective term of the query and marking each respective term as either appropriate for the first age group or inappropriate for the first age group.
14. The system of claim 9 , wherein the control circuitry is configured to determine the inputted term of the query is inappropriate for the first age group by:
determining that the inputted term matches a term in a list of terms marked as inappropriate for the first age group in the identified context.
15. The system of claim 14 , wherein the list of terms marked as inappropriate for the first age group in the identified context comprises a list of commonly misused terms by users in the first age group in the identified context.
16. The system of claim 14 , wherein the list of terms marked as inappropriate for the first age group in the identified context comprises a list of commonly mispronounced terms by users in the first age group in the identified context.