Evaluating user status via natural language processing and machine learning

Patent metadata

Field	Value
Publication number	US-12561684-B2
Application number	US-202418630871-A
Country	US
Kind code	B2
Filing date	Apr 9, 2024
Priority date	May 26, 2020
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

A request is received from a first user of a social interaction platform. The request is a request to acquire a status. In response to the receiving of the request, a database is accessed. The database contains first electronic data pertaining to previous interactions between the first user and other entities of the social interaction platform. The first electronic data is analyzed via one or more Natural Language Processing (NLP) techniques. A first result is obtained based on the analyzing. A machine learning process is executed based at least in part on the first result. Based on the executing of the machine learning process, a determination is made whether to grant or deny the request received from the first user.

First claim

What is claimed is:

1 . A method, comprising:
accessing a language pattern of one or more reference users of an electronic interaction platform, wherein the language pattern was generated via one or more Natural Language Processing (NLP) techniques, and wherein the one or more reference users have acquired a specified status; training, using the language pattern of the one or more reference users as training data, a machine learning model; accessing electronic data associated with a target user on the electronic interaction platform, wherein the target user comprises a user who does not have the specified status; and determining, at least in part via a prediction made by the trained machine learning model based on the electronic data, whether to grant the specified status to the target user; wherein the accessing, the training, the accessing, or the determining is performed using a system that includes one or more hardware processors.

2 . The method of claim 1 , wherein the electronic interaction platform is configured to allow a plurality of users to electronically post and view content, wherein the plurality of the users include the one or more reference users and the target user.

3 . The method of claim 2 , wherein the language pattern is associated with the content posted by the one or more reference users.

4 . The method of claim 2 , wherein the content posted by the one or more reference users include graphical but non-textual content.

5 . The method of claim 1 , further comprising generating, via the one or more NLP techniques and based on the electronic data of the target user, a language pattern of the target user, wherein the prediction is obtained at least in part by inputting the language pattern of the target user to the machine learning model.

6 . The method of claim 1 , wherein the training comprises training a classification tree model or a regression tree model.

7 . The method of claim 1 , wherein the specified status is associated with a decision to grant or deny a loan or a credit.

8 . The method of claim 1 , wherein the language pattern of the one or more reference users is generated based on a same type of data as the electronic data associated with the target user.

9 . The method of claim 1 , wherein the prediction includes a score, and wherein the determining is based on whether the score meets a threshold score.

10 . The method of claim 1 , wherein the NLP technique comprises a technique to vectorize a plurality of words in the language pattern.

11 . The method of claim 1 , wherein the NLP technique generates a numerical statistic that reflects a relative importance of one or more words in the language pattern.

12 . A system, comprising:
a natural language processing (NLP) module configured to perform operations that include:
accessing first electronic data of one or more reference users of an electronic interaction platform, wherein the electronic interaction platform facilitates electronic communications among a plurality of users that include the one or more reference users, and wherein the one or more reference users have been granted a specified status;
accessing second electronic data of an applicant user of the plurality of users of the electronic interaction platform, wherein the applicant user is applying for the specified status;
generating a first NLP analysis result based on the first electronic data for the one or more references user; and
generating a second NLP analysis result based on the second electronic data for the applicant user;
a machine learning module configured to perform operations that include:
training a machine learning model at least in part by using the first NLP analysis result as training data; and
generating an output at least in part by inputting the second NLP analysis result into the trained machine learning model, wherein the output indicates a degree of similarity between the first electronic data and the second electronic data; and
a decision module configured to determine, based at least in part on the output generated by the machine learning module, whether to grant or deny the specified status for the applicant user.

13 . The system of claim 12 , wherein the first electronic data or the second electronic data comprises an electronic message, an electronic memorandum, an electronic chat record, an electronic status update, or an electronic record of a transaction.

14 . The system of claim 12 , wherein:
the first NLP analysis result is associated with a first language pattern of the one or more reference users; and the second NLP analysis result is associated with a second language pattern of the applicant user.

15 . The system of claim 14 , wherein the output comprises a score that indicates a degree of similarity between the first language pattern and the second language pattern.

16 . The system of claim 12 , wherein the first electronic data or the second electronic data contains an audio or an image.

17 . The system of claim 12 , further comprising a meta data scoring module configured to generate a score for the applicant user based on meta data associated with the applicant user, the meta data indicating a type of the specified status, a history of the applicant user with the electronic interaction platform, or a transaction record of the applicant user, wherein a determination of whether to grant or deny the specified status for the applicant user from the decision module is based further on the score generated by the meta data scoring module.

18 . A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:
performing a first Natural Language Processing (NLP) process to first electronic interaction data generated by one or more first users of an online platform, wherein the one or more first users have been granted a specified status; training, using a first result of the first NLP process as training data, a machine learning model; performing a second NLP process to second electronic interaction data generated by a second user of the online platform, wherein the second user has requested, but has not been granted, the specified status, and wherein the first electronic interaction data and the second electronic interaction data each contain a plurality of words of a human language; determining, at least in part by inputting a result of the second NLP process to the machine learning model, a similarity between the second user and at least one of the one or more first users; and granting or denying the specified status to the second user based at least in part on the determining.

19 . The non-transitory machine-readable medium of claim 18 , wherein the first electronic interaction data or the second electronic interaction data further contains non-textual data.

20 . The non-transitory machine-readable medium of claim 18 , wherein:
the first electronic interaction data or the second electronic interaction data is associated with one or more electronic postings or transaction records of the one or more first users or the second user; and the first NLP process or the second NLP process comprises a word2vec process performed on the one or more electronic postings or the transaction records.

Assignees

• Paypal Inc

Inventors

• Hochma Gal
• Rotenberg Matias
• Cohen Yael
• Yuchtman Ran
• Elbaz Shay
• Levkovich Chen

Classifications

• G06Q40/03

  Credit; Loans; Processing thereof · CPC title

• G06F40/35

  Discourse or dialogue representation · CPC title

• G06Q20/384

  using social networks · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title