Systems and methods for search query refinement

A system comprising:

• one or more processors;
• and
• one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to and perform:
• determining a set of questions associated with a campaign by extracting text from one or more advertisements;
• storing one or more previous responses to a first question of the set of questions in a user profile of a user;
• training a deep neural network model using historical data for one or more online activities as input vector data to the deep neural network model to predict a probability associated with each of the set of questions as output;
• generating vector embeddings of the one or more online activities of the user as input to the deep neural network model, wherein an output of the deep neural network model comprises a respective relevance probability associated with different questions in the set of questions based on the one or more online activities of the user;
• updating, the deep neural network model at least in part on feedback from the user by tracking engagement data from the user based on at least one of the set of questions, wherein the engagement data is used in a feedback loop to tune the deep neural network model for one or more users with online activities similar to the one or more online activities of the user;
• receiving a search query from a graphical user interface of a computing device of the user;
• determining a second question from the set of questions to present to the user based on the respective relevance probability associated with the different questions in the set of questions;
• evaluating one or more previous responses stored in the user profile of the user for the search query to prevent presenting the second question from the set of questions more than once to the user in response to the search query;
• conducting an evaluation of one or more user actions during a current browse session of the user associated with the second question;
• determining a confidence score associated with the second question;
• and
• when the confidence score associated with the second question exceeds a second predetermined threshold, presenting the second question to the user via the graphical user interface in response to receiving the search query.