System and method for presenting targeted content

A computer-implemented method comprising:

• obtaining, from a user device, user permission and accessing profile attributes and friend attributes from a social-network interface;
• generating for each person, an identifier from at least name, date of birth and location;
• loading, keyed by the identifier, user activity data associated with the identifier from a database, the user activity data comprising (i) search and browsing actions of the user and (ii) external records ingested via a distributed data pipeline;
• determining, using a classifier trained on features including search keywords, domains, and page categories, a purchase-funnel stage associated with the identifier, the purchase-funnel stage identifying at least one of a discovery stage, a consideration stage, and an action stage;
• generating a first advertisement message corresponding to the determined purchase-funnel stage;
• displaying, on the user device during an interaction session, the first advertisement message as a part of conversation dialogues;
• storing the conversational dialogues in both (i) a recurrent neural network as questions and answers using sequence-to-sequence algorithms, and (ii) an inverted-index graph data structure as parent nodes representing user dialogues and corresponding child nodes representing response dialogues;
• loading updated interaction data generated in response to user input received during the interaction session;
• applying a purchase-funnel progression model to the updated interaction data to determine whether criteria for advancing the purchase-funnel stage are satisfied;
• in response to a determination that the criteria are satisfied, generating an updated advertisement message corresponding to a different purchase-funnel stage;
• delivering, on the user device, the updated advertisement message, for presentation on the user device, and concurrently inserting the updated advertisement message as a sequence-to-sequence query into the recurrent neural network and as a new parent node into the inverted-index graph data structure;
• in response to user interaction with the updated advertisement message during the interaction session, generating, by a virtual agent server, an order request associated with the updated advertisement message;
• and
• storing in a database, each user-advertisement interaction keyed by the identifier.

A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause the processors to perform operations comprising:

• obtaining user permission and accessing profile attributes and friend attributes from a social-network interface;
• generating for each person, an identifier from at least name, date of birth and location;
• loading, keyed by the identifier, user activity data associated with the identifier from a database, the user activity data comprising (i) search and browsing actions of the user and (ii) external records ingested via a distributed data pipeline;
• determining, using a classifier trained on features including search keywords, domains, and page categories, a purchase-funnel stage associated with the identifier, the purchase-funnel stage identifying at least one of a discovery stage, a consideration stage, and an action stage;
• generating a first advertisement message corresponding to the determined purchase-funnel stage;
• displaying, on the user device during an interaction session, the first advertisement message as a part of conversation dialogues;
• storing the conversational dialogues in both (i) a recurrent neural network as questions and answers using sequence-to-sequence algorithms, and (ii) an inverted-index graph data structure as parent nodes representing user dialogues and corresponding child nodes representing response dialogues;
• loading updated interaction data generated in response to user input received during the interaction session;
• applying a purchase-funnel progression model to the updated interaction data to determine whether criteria for advancing the purchase-funnel stage are satisfied;
• in response to a determination that the criteria are satisfied, generating an updated advertisement message corresponding to a different purchase-funnel stage;
• delivering, on the user device, the updated advertisement message, for presentation on the user device, and concurrently inserting the updated advertisement message as a sequence-to-sequence query into the recurrent neural network and as a new parent node into the inverted-index graph data structure;
• in response to user interaction with the updated advertisement message during the interaction session, generating, by a virtual agent server, an order request associated with the updated advertisement message;
• and
• storing each user-advertisement interaction keyed by the identifier.