Contextual natural language censoring
Systems and methods for contextual natural language censoring are disclosed. For example, configuration data indicating details associated with content provided by a client device and/or about the client device may be received and may be utilized to determine impermissible and permissible exceptions for a given client. One or more queries may be generated utilizing the impermissible and permissible exceptions, and when input data is received from the client device and/or in association with the client identifier, the queries may be utilized to evaluate the input data for impermissible and permissible exceptions. The results may be filtered based on user preferences, the input data may be censored, the input data may be prevented from being exposed to a user device, the application associated with the input data may be removed from availability, and/or a maturity setting may be changed for the application, for example.
1. A method comprising:
receiving first data representing a user request for content generated by an application;
sending the first data to the application;
receiving second data representing the content from the application;
generating third data indicating a policy violation schema, the third data identifying:
a first phrase indicated to be impermissible; and
a second phrase indicated to be a permissible exception based at least in part on a contextual meaning, the second phrase encompassing the first phrase;
determining, based at least in part on the third data, that the second data includes the first phrase;
determining a policy violation severity associated with the second data including the first phrase, wherein the policy violation severity is based on the application;
selecting a corrective action to perform on the content generated by the application from multiple corrective actions based at least in part on the policy violation severity, the corrective action including sending a request for the application to generate altered content;
generating the request for the application to generate the altered content based at least in part on selecting the corrective action;
sending the request to the application; and
in response to receiving the altered content from the application, causing a selected device to output of the altered content.
2. The method of claim 1 , further comprising:
determining permissible exceptions, including the permissible exception, associated with the contextual meaning; and
generating the third data based at least in part on the permissible exceptions.
3. The method of claim 1 , further comprising:
receiving fourth data representing a request to validate text data associated with the application;
receiving the text data;
determining that impermissible phrases are absent from the text data; and
outputting a response indicating that the impermissible phrases are absent from the text data.
4. The method of claim 1 , further comprising:
determining that the first phrase is associated with a phrase type indicated to be violative of a content policy;
determining, based at least in part on the first phrase being associated with the phrase type, a content identifier associated with a portion of the application; and
causing the content identifier to be removed from an application list indicating portions of the application to be utilized.
5. The method of claim 1 , wherein the application is configured to perform the corrective action by generating the altered content that excludes the first phrase.
6. The method of claim 1 , further comprising generating, based at least in part on the first data, a request for the content to be generated by the application, the request comprising a natural language representation of the user request along with a command to produce the content.
7. The method of claim 1 , further comprising:
storing fourth data indicating:
impermissible phrases associated with an input type; and
permissible exceptions, including the permissible exception, associated with the input type;
determining that the application is associated with the input type; and
generating the third data based at least in part on the application being associated with the input type.
8. The method of claim 1 , further comprising:
storing fourth data indicating:
impermissible phrases associated with a language of multiple languages; and
permissible exceptions, including the permissible exception, associated with the language;
determining that the application is associated with the language; and
generating the third data based at least in part on the application being associated with the language.
9. The method of claim 1 , wherein:
the application is associated with a language model configured to generate the content in response to user requests; and
sending the first data comprises sending text data representing the first data to the application.
10. The method of claim 1 , wherein the user request comprises audio data, and the method further comprises:
generating, utilizing the audio data, text data representing the user request;
generating a command for the application to provide the content utilizing the text data; and
receiving the altered content from the application.
11. A system comprising:
one or more processors; and
non-transitory computer-readable media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
receiving first data representing a user request for content generated by an application;
sending the first data to the application;
receiving second data representing the content from the application;
generating third data indicating a policy violation schema, the third data identifying:
a first phrase indicated to be impermissible; and
a second phrase indicated to be a permissible exception based at least in part on a contextual meaning, the second phrase encompassing the first phrase;
determining, based at least in part on the third data, that the second data includes the first phrase;
determining a policy violation severity associated with the second data including the first phrase, wherein the policy violation severity is based on the application;
selecting a corrective action to perform on the content generated by the application from multiple corrective actions based at least in part on the policy violation severity, the corrective action including sending a request for the application to generate altered content;
generating the request for the application to generate the altered content based at least in part on selecting the corrective action;
sending the request to the application; and
in response to receiving the altered content from the application, causing a selected device to output of the altered content.
12. The system of claim 11 , the operations further comprising:
determining permissible exceptions, including the permissible exception, associated with the contextual meaning; and
generating the third data based at least in part on the permissible exceptions.
13. The system of claim 11 , the operations further comprising:
receiving fourth data representing a request to validate text data associated with the application;
receiving the text data;
determining that impermissible phrases are absent from the text data; and
outputting a response indicating that the impermissible phrases are absent from the text data.
14. The system of claim 11 , the operations further comprising:
determining that the first phrase is associated with a phrase type indicated to be violative of a content policy;
determining, based at least in part on the first phrase being associated with the phrase type, a content identifier associated with a portion of the application; and
causing the content identifier to be removed from an application list indicating portions of the application to be utilized.
15. The system of claim 11 , wherein the application is configured to perform the corrective action by generating the altered content that excludes the first phrase.
16. The system of claim 11 , the operations further comprising generating, based at least in part on the first data, a request for the content to be generated by the application, the request comprising a natural language representation of the user request along with a command to produce the content.
17. The system of claim 11 , the operations further comprising:
storing fourth data indicating:
impermissible phrases associated with an input type; and
permissible exceptions, including the permissible exception, associated with the input type;
determining that the application is associated with the input type; and
generating the third data based at least in part on the application being associated with the input type.
18. The system of claim 11 , the operations further comprising:
storing fourth data indicating:
impermissible phrases associated with a language of multiple languages; and
permissible exceptions, including the permissible exception, associated with the language;
determining that the application is associated with the language; and
generating the third data based at least in part on the application being associated with the language.
19. The system of claim 11 , wherein:
the application is associated with a language model configured to generate the content in response to user requests; and
sending the first data comprises sending text data representing the first data to the application.
20. The system of claim 11 , wherein the user request comprises audio data, and the operations further comprise:
generating, utilizing the audio data, text data representing the user request;
generating a command for the application to provide the content utilizing the text data; and
receiving the altered content from the application.