User conversion prediction using a multi-task model

A computer system, comprising:

• at least one processor;
• and
• at least one memory storing instructions that, when executed by the at least processor, configure the computing system to perform operations comprising:
• generating a user embedding representation by processing user interaction data through a first series of neural network layers of a neural network;
• generating an advertisement embedding representation by processing advertisement features through a second series of neural network layers of the neural network;
• determining a task representation based on the user embedding representation and the advertisement embedding representation;
• applying user personalized data and the task representation to a parameter personalized network (PPNET), the PPNET comprising a gating network and task-specific expert networks, the PPNET configured to apply the task representation to a gating network of the PPNET in order to control, by the gating network, information among the task-specific expert networks;
• calculating, using the PPNET, probabilities of a user action, the user action comprising a user interaction with an advertisement and at least one of page-view, sign-up, add-to-cart, and purchase after the user interaction with the advertisement;
• calculating a click-through conversion (CTC) probability based on the probabilities of one or more user actions;
• ranking one or more content items based on the CTC probability;
• and
• causing display of the one or more content items based on the ranking.

A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to perform operations comprising:

• generating a user embedding representation by processing user interaction data through a first series of neural network layers of a neural network;
• generating an advertisement embedding representation by processing advertisement features through a second series of neural network layers of the neural network;
• determining a task representation based on the user embedding representation and the advertisement embedding representation;
• applying user personalized data and the task representation to a parameter personalized network (PPNET), the PPNET comprising a gating network and task-specific expert networks, the PPNET configured to apply the task representation to a gating network of the PPNET in order to control, by the gating network, information among the task-specific expert networks;
• calculating, using the PPNET, probabilities of a user action, the user action comprising a user interaction with an advertisement and at least one of page-view, sign-up, add-to-cart, and purchase after the user interaction with the advertisement;
• calculating a click-through conversion (CTC) probability based on the probabilities of one or more user actions;
• ranking one or more content items based on the CTC probability;
• and
• causing display of the one or more content items based on the ranking.