Generating query variants using a trained generative model
Systems, methods, and computer readable media related to generating query variants for a submitted query. In many implementations, the query variants are generated utilizing a generative model. A generative model is productive, in that it can be utilized to actively generate a variant of a query based on application of tokens of the query to the generative model, and optionally based on application of additional input features to the generative model.
1 . A method implemented by one or more processors, the method comprising:
receiving an original query, the original query generated based on user interface input of a user via a client device;
obtaining, from a search system, one or more search system responses to the original query;
applying, to a trained generative model:
one or more attributes of the user that are determined based on one or more query submissions received prior to the receiving of the original query,
the original query, and
the one or more search system responses to the original query;
generating, based on application of the original query, application of the one or more search system responses, and application of the one or more attributes of the user, to the trained generative model, generative output;
subsequent to generating the generative output:
generating, based on application of further input to the trained generative model, further generative output,
wherein the further input is based on the generative output;
generating an output based on the further generative output; and
providing, in response to the original query, the output for presentation via the client device.
2 . The method of claim 1 , wherein the one or more attributes of the user, applied to the trained generative model, comprise:
a location of the user,
a task currently engaged in by the user, and/or
weather at the location of the user.
3 . The method of claim 1 , further comprising:
applying, along with the original query, the one or more search system responses, and the one or more attributes of the user, to the trained generative model:
content recently viewed by the user at the client device;
wherein generating the generative output is further based on application of the content to the trained generative model.
4 . The method of claim 1 , further comprising:
selecting the trained generative model from a plurality of candidate generative models;
wherein applying the original query, the one or more search system responses, and the one or more attributes of the user, to the trained generative model is in response to selecting the trained generative model from the plurality of candidate generative models.
5 . The method of claim 4 , wherein selecting the trained generative model from the plurality of candidate generative models is based on the one or more user attributes of the user that generated the original query.
6 . The method of claim 4 , wherein selecting the trained generative model from the plurality of candidate generative models is based on a task currently engaged in by the user.
7 . The method of claim 1 , wherein the further input includes the generative output.
8 . The method of claim 1 , wherein the further input includes further search system responses that are responsive to the generative output.
9 . A system comprising:
one or more storage devices storing instructions; and
one or more processors operable to execute the instructions to:
receive an original query, the original query generated based on user interface input of a user via a client device;
obtain, from a search system, one or more search system responses to the original query;
apply, to a trained generative model:
one or more attributes of the user that are determined based on one or more query submissions received prior to the receiving of the original query,
the original query, and
the one or more search system responses to the original query;
generate, based on application of the original query, application of the one or more search system responses, and application of the one or more attributes of the user, to the trained generative model, generative output;
subsequent to generating the generative output:
generate, based on application of further input to the trained generative model, further generative output,
wherein the further input is based on the generative output;
generate an output based on the further generative output; and
provide, in response to the original query, the output for presentation via the client device.
10 . The system of claim 9 , wherein the one or more attributes of the user, applied to the trained generative model, comprise:
a location of the user,
a task currently engaged in by the user, and/or
weather at the location of the user.
11 . The system of claim 9 , wherein at least a processor of the one or more processors is further operable to execute the instructions to:
apply, along with the original query, the one or more search system responses, and the one or more attributes of the user, to the trained generative model:
content recently viewed by the user at the client device;
wherein generating the generative output is further based on application of the content to the trained generative model.
12 . The system of claim 9 , wherein at least a processor of the one or more processors is further operable to execute the instructions to:
select the trained generative model from a plurality of candidate generative models;
wherein applying the original query, the one or more search system responses, and the one or more attributes of the user, to the trained generative model is in response to selecting the trained generative model from the plurality of candidate generative models.
13 . The system of claim 12 , wherein selecting the trained generative model from the plurality of candidate generative models is based on the one or more user attributes of the user that generated the original query.
14 . The system of claim 12 , wherein selecting the trained generative model from the plurality of candidate generative models is based on a task currently engaged in by the user.
15 . A non-transitory computer-readable medium storing software comprising instructions executable by one or more processors to:
receive an original query, the original query generated based on user interface input of a user via a client device;
obtain, from a search system, one or more search system responses to the original query;
apply, to a trained generative model:
one or more attributes of the user that are determined based on one or more query submissions received prior to the receiving of the original query,
the original query, and
the one or more search system responses to the original query;
generate, based on application of the original query, application of the one or more search system responses, and application of the one or more attributes of the user, to the trained generative model, generative output;
subsequent to generating the generative output:
generate, based on application of further input to the trained generative model, further generative output,
wherein the further input is based on the generative output;
generate an output based on the further generative output; and
provide, in response to the original query, the output for presentation via the client device.
16 . The non-transitory computer-readable medium of claim 15 , wherein the one or more attributes of the user comprise:
a location of the user,
a task currently engaged in by the user, and/or
weather at the location of the user;
wherein generating the generative output is further based on application of the location, the task, and/or the weather to the trained generative model.
17 . The non-transitory computer-readable medium of claim 15 , wherein the instructions are further executable by at least a processor of the one or more processors to:
apply, along with the original query and the one or more search system responses, and the one or more attributes of the user, to the trained generative model:
content recently viewed by the user at the client device;
wherein generating the generative output is further based on application of the content to the trained generative model.
18 . The non-transitory computer-readable medium of claim 15 , wherein the instructions are further executable by at least a processor of the one or more processors to:
select the trained generative model from a plurality of candidate generative models;
wherein applying the original query, the one or more search system responses, and the one or more attributes of the user, to the trained generative model is in response to selecting the trained generative model from the plurality of candidate generative models.
19 . The non-transitory computer-readable medium of claim 18 , wherein selecting the trained generative model from the plurality of candidate generative models is based on the one or more user attributes of the user that generated the original query.
20 . The non-transitory computer-readable medium of claim 18 , wherein selecting the trained generative model from the plurality of candidate generative models is based on a task currently engaged in by the user.