REMOVING PERSONAL INFORMATION FROM TEXT USING A NEURAL NETWORK

• US 20190332658A1
• Filed: 11/13/2018
• Published: 10/31/2019
• Est. Priority Date: 04/27/2018
• Status: Active Grant

• Current Assignee: Asapp, Inc.
• Original Assignee: Asapp, Inc.
• Inventors: Heckel, Frederick William Poe, Henry, Shawn
• Granted Patent: US 10,878,181 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 21/60 Protecting data
• CPC Class Codes: G06F 21/6245 Protecting personal data, e.g. for financial or medical purposes
• CPC Class Codes: G06F 40/166 Editing, e.g. inserting or deleting
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/048 Activation functions
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A neural network may be used to remove personal information from text (such as names, addresses, credit card numbers, or social security numbers), and replace the personal information with a label indicating the type or class of the removed information. The neural network may comprise multiple layers that compute a context vector for words of the text, compute label scores for words of the text using the context vectors, and select a label for each word using the label scores. Words corresponding to certain labels may be replaced with a label, such as replacing the digits of a credit card number with a label <cc_number>. The redacted text may then be presented to a person or stored for later processing.

Claims

1. A computer-implemented method for removing personal information from text, comprising:

receiving text of a message from a user;

generating first redacted text by processing the text with a first level of redaction;

storing the first redacted text in in-memory storage;

presenting the first redacted text to a customer service representative who is assisting the user;

removing the first redacted text from the in-memory storage after a first period of time;

generating second redacted text by processing the text or the first redacted text with a second level of redaction, wherein;

the second level of redaction is higher than the first level of redaction, and

the second level of redaction replaces redacted text with a label indicating a type of information that was removed from the text;

storing the second redacted text in durable storage; and

removing the second redacted text from the durable storage after a second period of time, wherein the second period of time is longer than the first period of time.

2. The computer-implemented method of claim 1, wherein the second level of redaction is performed using a neural network.

3. The computer-implemented method of claim 2, wherein the second level of redaction comprises computing a context vector for each word of the text.

4. The computer-implemented method of claim 1, wherein the in-memory storage comprises a key-value store.

5. The computer-implemented method of claim 1, wherein the durable storage comprises a database stored to disk.

6. The computer-implemented method of claim 1, further comprising receiving the text of the message from the user as part of a customer support session between the user and the customer service representative, and wherein the first redacted text is removed from the in-memory storage after conclusion of the customer support session.

7. The computer-implemented method of claim 1, wherein the second level of redaction removes at least one of a phone number, a mailing address, an email address, or a name from the text.

8. The computer-implemented method of claim 1, comprising:

generating third redacted text by processing the text, the first redacted text, or the second redacted text with a third level of redaction, wherein;

the third level of redaction is higher than the second level of redaction, and

the third level of redaction replaces redacted text with a label indicating a type of information that was removed from the text;

storing the third redacted text in second durable storage, wherein the third redacted text is stored for longer than the second period of time.

9. A system, comprising:

at least one computer comprising at least one processor and at least one memory, the at least one computer configured to;

receive text of a message from a user;

generate first redacted text by processing the text with a first level of redaction;

store the first redacted text in in-memory storage;

present the first redacted text to a customer service representative who is assisting the user;

remove the first redacted text from the in-memory storage after a first period of time;

generate second redacted text by processing the text or the first redacted text with a second level of redaction, wherein the second level of redaction is higher than the first level of redaction;

store the second redacted text in durable storage;

remove the second redacted text from the durable storage after a second period of time, wherein the second period of time is longer than the first period of time;

generate third redacted text by processing the text, the first redacted text, or the second redacted text with a third level of redaction, wherein the third level of redaction is higher than the second level of redaction; and

store the third redacted text in second durable storage, wherein the third redacted text is stored for longer than the second period of time.

10. The system of claim 9, wherein the third level of redaction removes all personal information from the text.

11. The system of claim 9, wherein the third level of redaction removes all numbers from the text.

12. The system of claim 9, wherein the third redacted text is used for training models or classifiers.

13. The system of claim 9, wherein:

the second level of redaction replaces redacted text with a label indicating a type of information that was removed from the text; and

the third level of redaction replaces redacted text with a label indicating a type of information that was removed from the text.

14. The system of claim 9, wherein the second redacted text is used for logging of customer support.

15. One or more non-transitory, computer-readable media comprising computer executable instructions that, when executed, cause at least one processor to perform actions comprising:

receiving text of a message from a user;

generating first redacted text by processing the text with a first level of redaction;

storing the first redacted text in in-memory storage;

presenting the first redacted text to a customer service representative who is assisting the user;

removing the first redacted text from the in-memory storage after a first period of time;

generating second redacted text by processing the text or the first redacted text with a second level of redaction, wherein;

the second level of redaction is higher than the first level of redaction, and

the second level of redaction replaces redacted text with a label indicating a type of information that was removed from the text;

storing the second redacted text in durable storage; and

removing the second redacted text from the durable storage after a second period of time, wherein the second period of time is longer than the first period of time.

16. The one or more non-transitory, computer-readable media of claim 15, wherein the actions comprise:

generating third redacted text by processing the text, the first redacted text, or the second redacted text with a third level of redaction, wherein;

the third level of redaction is higher than the second level of redaction, and

the third level of redaction replaces redacted text with a label indicating a type of information that was removed from the text; and

storing the third redacted text in second durable storage, wherein the third redacted text is stored for longer than the second period of time.

17. The one or more non-transitory, computer-readable media of claim 15, wherein the first level of redaction is performed using a regular expression.

18. The one or more non-transitory, computer-readable media of claim 15, wherein the first level of redaction removes at least one of a credit card number or a social security number from the text.

19. The one or more non-transitory, computer-readable media of claim 15, wherein the first period of time is less than one hour.

20. The one or more non-transitory, computer-readable media of claim 15, wherein the second period of time is longer than one hour and less than one year.