Machine-learning-based processing of de-obfuscated data for data enrichment

Patent metadata

Field	Value
Publication number	US-2019102574-A1
Application number	US-201816007787-A
Country	US
Kind code	A1
Filing date	Jun 13, 2018
Priority date	Sep 29, 2017
Publication date	Apr 4, 2019
Grant date	—

Abstract

Embodiments relate to de-obfuscated data to process at one or more artificial-intelligence (AI) decision nodes throughout a communication decision tree. Through processing directed by the communication decision tree, additional de-obfuscated data is collected and securely transmitted in de-obfuscated or re-obfuscated form to enrich underlying data structures.

First claim

What is claimed is:

1 . A computer-implemented method comprising:
receiving, from a client system, a request to perform a machine-learning communication workflow; receiving, in association with the request, a set of obfuscated identifiers for which processing via the machine-learning communication workflow is requested, wherein each obfuscated identifiers of the set of obfuscated identifiers corresponds to an identification of an obfuscated version of a profile stored at a data management system, the obfuscated version lacking personally identifiable information (PII); for each obfuscated identifier in the set of obfuscated identifiers:
mapping the obfuscated identifier to a non-obfuscated identifier that identifies a non-obfuscated version of the profile that includes PII;
retrieving, from the data management system, user data from the non-obfuscated version;
retrieving learned data generated by training a machine-learning model using other user data;
executing the machine-learning model configured with the learned data to process at least part of the user data;
identifying one or more communication specifications based on the execution of the machine-learning model configured with the learned data;
causing content to be transmitted to a destination address identified in the user data in accordance with the one or more communication specifications; and
generating non-obfuscated communication-activity data for the non-obfuscated profile identifier based on any communications detected in response to the content transmission;
obfuscating the non-obfuscated communication-activity data to generate a set of obfuscated data for the set of obfuscated identifiers; transmitting at least part of the set of obfuscated data to the client system; and causing, for each of at least some of the set of obfuscated identifiers, at least some of the non-obfuscated communication-activity data to be stored in association with the non-obfuscated profile identifier to the data management system.

2 . The method of claim 1 , further comprising:
accessing a data structure representing a communication decision tree configured to dynamically define individual trajectories through the communication decision tree using a machine-learning technique to indicate a series of communication specifications, the communication decision tree including a set of branching nodes, each branching node of the set of branching nodes corresponding to an action point configured to identify a direction for a given trajectory; and for each obfuscated identifier in the set of obfuscated identifiers, prior to executing the machine-learning model:
initiating a trajectory;
detecting that the trajectory has reached a particular branching node of the set of branching nodes, the particular branching node being connected to multiple other nodes, each of the next nodes corresponding to at least one communication specification;
wherein a result of the execution of the machine-learning model configured with the learning data corresponds to a selection of a particular other node of the multiple other nodes, the at least one communication specification corresponding to the particular next node including the identified one or more communication specifications.

3 . The method of claim 2 , further comprising, for each obfuscated identifier in the set of obfuscated identifiers:
determining, based at least in part on the any communications detected in response to the content transmission, whether and/or how to advance the trajectory; when it is determined that the trajectory is to be advanced, advancing the trajectory; identifying a last node of the trajectory; and assigning the trajectory to a communication stage of a set of communication stages based on the last node, wherein the at least part of the set of obfuscated data includes an identification of a particular communication stage of the set of communication stages and a subset of the set of obfuscated identifiers, each obfuscated identifier in the subset having had a trajectory assigned to the particular communication stage.

4 . The method of claim 2 , further comprising, for each obfuscated identifier of at least some of the set of obfuscated identifiers:
detecting a request communication responsive to the content transmission, the request communication requesting other content; advancing the trajectory to another particular branching node of the set of branching nodes; retrieving other learned data generated by training another iteration of the machine-learning model; executing the other iteration of the machine-learning model configured with the other learned data to process another at least part of the user data; identifying one or more other communication specifications based on the execution of the other iteration of the machine-learning model configured with the other learned data; and causing other content to be transmitted to the destination address in accordance with the one or more other communication specifications, wherein the any communications detected in response to the content transmission includes any communication detected in response to the other content transmission.

5 . The method of claim 1 , further comprising, for each obfuscated identifier of at least some of the set of obfuscated identifiers:
receiving one or more new communications, each of the one or more new communications having been transmitted from a user device in response to the content transmission or to another content transmission transmitted as part of the machine-learning communication workflow; and extracting field values from the one or more new communications, wherein the at least part of the set of obfuscated data includes the extracted field values in association with the obfuscated identifier.

6 . The method of claim 5 , further comprising:
receiving one or more indications as to which of the extracted field values added information to the obfuscated versions of the profiles associated with the set of obfuscated identifiers; and training an implementation of the machine-learning model using the one or more indications, the implementation of the machine-learning model being configured to prioritize informative data collection.

7 . The method of claim 1 , wherein the learned data includes one or more parameters corresponding to a regression model, and wherein the machine-learning model includes the regression model.

8 . A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform operations including:
receiving, from a client system, a request to perform a machine-learning communication workflow; receiving, in association with the request, a set of obfuscated identifiers for which processing via the machine-learning communication workflow is requested, wherein each obfuscated identifiers of the set of obfuscated identifiers corresponds to an identification of an obfuscated version of a profile stored at a data management system, the obfuscated version lacking personally identifiable information (PII); for each obfuscated identifier in the set of obfuscated identifiers:
mapping the obfuscated identifier to a non-obfuscated identifier that identifies a non-obfuscated version of the profile that includes PII;
retrieving, from the data management system, user data from the non-obfuscated version;
retrieving learned data generated by training a machine-learning model using other user data;
executing the machine-learning model configured with the learned data to process at least part of the user data;
identifying one or more communication specifications based on the execution of the machine-lear

Assignees

• Oracle Int Corp

Inventors

• Roberts Tara U
• Stanesby Jonathan
• Pearmain Michael
• Yordy Kevin
• Sutthakorn Karen

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N20/20 Primary

  Ensemble learning · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N5/025

  Extracting rules from data · CPC title