Systems and methods for generating a plain English interpretation of a legal clause

• US 10,628,530 B1
• Filed: 07/18/2019
• Issued: 04/21/2020
• Est. Priority Date: 12/07/2018
• Status: Active Grant

• Current Assignee: Capital One Services LLC (Capital One Financial Corporation)
• Original Assignee: Capital One Services LLC (Capital One Financial Corporation)
• Inventors: Walters, Austin, Goodsitt, Jeremy Edward, Abad, Fardin Abdi Taghi, Farivar, Reza, Pham, Vincent, Watson, Mark, Taylor, Kenneth, Truong, Anh
• Primary Examiner(s): Patel, Shreyans A
• Application Number:
  US16/515,650
• Time in Patent Office: 278 Days
• CPC Class Codes: G06F 18/2178 based on feedback of a supervisor
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation
• CPC Class Codes: G06F 40/44 Statistical methods, e.g. probability models
• CPC Class Codes: G06F 40/45 Example-based machine translation; Alignment
• CPC Class Codes: G06F 40/56 Natural language generation
• 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 3/08 Learning methods
• CPC Class Codes: G06Q 50/18 Legal services
• CPC Class Codes: G06V 10/82 using neural networks
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
• CPC Class Codes: H04L 51/046 Interoperability with other network applications or services
• CPC Class Codes: H04L 51/08 Annexed information, e.g. attachments

Abstract

A system is configured to perform one or more steps of a method. The system may receive a plurality of attorney communications, identify one or more legal clause interpretations in them, receive a first legal clause and provide it to a trained NN and a probability model. The system may also generate a corresponding first plain English interpretation based on the first legal clause, provide the first plain English interpretation to the probability model, which generates a probability score based on a degree to which the legal clause matches the plain English interpretation in meaning, and determine whether the probability score exceeds a predetermined threshold. Further, the system may instruct the NN to generate a second plain English interpretation based on the first legal clause when the probability score does not exceed the predetermined threshold, and output the first plain English interpretation when the probability score exceeds the predetermined threshold.

Claims

1. A system, comprising:

one or more processors; and

a memory in communication with the one or more processors and storing instructions that, when executed by the one or more processors, are configured to cause the system to;

identify one or more legal clause interpretations in a plurality of attorney communications;

train a neural network (NN) based on the identified one or more legal clause interpretations;

provide a first legal clause to the trained NN and a probability model;

generate, via the trained NN, a first non-legalese interpretation based on the first legal clause;

provide the first non-legalese interpretation to a probability model;

generate, using the probability model, a probability score based on a degree to which the first legal clause matches the non-legalese interpretation in meaning;

determine whether the probability score exceeds a predetermined threshold;

when the probability score does not exceed the predetermined threshold, instruct the NN to generate a second non-legalese interpretation based on the first legal clause; and

when the probability score exceeds the predetermined threshold, output the first non-legalese interpretation.

2. The system of claim 1, wherein the probability model is a convolutional neural network (CNN) and the NN is either a CNN or a recurrent neural network (RNN).

3. The system of claim 2, wherein the plurality of attorney communications comprises a plurality of email communications.

4. The system of claim 3, wherein identifying the one or more legal clause interpretations in the plurality of attorney communications comprises detecting a redline change in a document attached to one of the plurality of email communications and identifying a paragraph associated with the redline change as a first legal clause interpretation of the one or more legal clause interpretations.

5. The system of claim 3, wherein identifying the one or more legal clause interpretations in the plurality of attorney communications comprises detecting an addition in a document attached to one of the plurality of email communications and identifying a paragraph associated with the addition as a first legal clause interpretation of the one or more legal clause interpretations.

6. The system of claim 3, wherein identifying the one or more legal clause interpretations in the plurality of attorney communications comprises detecting a comment in a document attached to one of the plurality of email communications and identifying text within the comment as a first legal clause interpretation of the one or more legal clause interpretations.

7. The system of claim 1, wherein the instructions, when executed by the one or more processors, are further configured to cause the system to:

receive, from a user device, reinforcement feedback based on the first non-legalese interpretation; and

iteratively re-train the trained NN based on the received reinforcement feedback.

8. The system of claim 6, wherein the output of the first non-legalese interpretation is in a chat program accessible by the user device and a reinforcement feedback is provided from the user device via the chat program.

9. A system, comprising:

one or more processors; and

a memory in communication with the one or more processors and storing instructions that, when executed by the one or more processors, are configured to cause the system to;

provide a first legal clause to a trained neural network (NN);

generate, via the trained NN, a first non-legalese interpretation based on the first legal clause;

receive, from a user device, reinforcement feedback based on the first non-legalese interpretation; and

iteratively re-train the trained NN based on the received reinforcement feedback.

10. The system of claim 9, wherein the NN is either a convolutional neural network (CNN) or a recurrent neural network (RNN).

11. The system of claim 10, wherein the instructions, when executed by the one or more processors, are further configured to cause the system to:

identify one or more legal clause interpretations in a plurality of attorney communications;

train the neural network based on the identified one or more legal clause interpretations, and

wherein the plurality of attorney communications comprises a plurality of email communications.

12. The system of claim 11, wherein identifying the one or more legal clause interpretation request in the plurality of attorney communications comprises detecting a redline change in a document attached to one of the plurality of email communications and identifying a paragraph associated with the redline change as a first legal clause interpretation of the one or more legal clause interpretations.

13. The system of claim 11, wherein identifying the one or more legal clause interpretation request in the plurality of attorney communications comprises detecting an addition in a document attached to one of the plurality of email communications and identifying a paragraph associated with the addition as a first legal clause interpretation of the one or more legal clause interpretations.

14. The system of claim 11, wherein identifying the one or more legal clause interpretation request in the plurality of attorney communications comprises detecting a comment in a document attached to one of the plurality of email communications and identifying text within the comment as a first legal clause interpretation of the one or more legal clause interpretations.

15. The system of claim 9, wherein the output of the first non-legalese interpretation is in a chat program accessible by the user device and the reinforcement feedback is provided from the user device via the chat program.

16. A system, comprising:

one or more processors; and

a memory in communication with the one or more processors and storing instructions that, when executed by the one or more processors, are configured to cause the system to;

provide a first legal clause to a trained neural network (NN) and a probability model;

generate, via the trained NN, a first non-legalese interpretation based on the first legal clause;

provide the first non-legalese interpretation to a probability model;

generate, using the probability model, a probability score based on a degree to which the legal clause matches the non-legalese interpretation in meaning;

determine whether the probability score exceeds a predetermined threshold;

when the probability score does not exceed the predetermined threshold, instruct the NN to generate a second non-legalese interpretation based on the first legal clause; and

when the probability score exceeds the predetermined threshold, output the first non-legalese interpretation.

17. The system of claim 16, wherein the probability model is a convolutional neural network (CNN) and the neural network is at either a CNN or a recurrent neural network (RNN).

18. The system of claim 16, wherein the instructions, when executed by the one or more processors, are further configured to cause the system to:

receive, from a user device, reinforcement feedback based on the first non-legalese interpretation; and

iteratively re-train the trained NN based on the received reinforcement feedback.

19. The system of claim 18, wherein the output of the first non-legalese interpretation is in a chat program accessible by the user device and the reinforcement feedback is provided from the user device via the chat program.

20. The system of claim 16, wherein the first non-legalese interpretation comprises a first plain English interpretation.