Systems and methods of natural language processing and statistical analysis to identify matching categories

• US 9,691,085 B2
• Filed: 04/20/2016
• Issued: 06/27/2017
• Est. Priority Date: 04/30/2015
• Status: Active Grant

• Current Assignee: Visa International Service Association (Visa, Inc.)
• Original Assignee: Visa International Service Association (Visa, Inc.)
• Inventors: Scheidelman, Edward
• Primary Examiner(s): GUERRA-ERAZO, EDGAR X
• Application Number:
  US15/134,261
• Publication Number:
  US 20160321705A1
• Time in Patent Office: 433 Days
• Field of Search: 704 9, 704 10
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute

Abstract

Combining the natural language processing of product descriptions and statistical analysis of payment data to classify consumers based on products purchased and merchants based on products sold. Systems and methods use natural language processing techniques to interpret the descriptions of item level purchase data to classify products that have been purchased by customers into micro-categories. Statistical deviation methods are applied to the payment data to calculate normalized mean product cost, after removing outliers. After determining the micro-categories of the products purchased and the mean product cost of the purchased products, the system and methods classify consumers and merchants into categories based at least in part on the product micro-categories, mean costs, and relative volume of product types sold by merchants to predict which consumers are likely to purchase from which merchants.

Claims

1. A method, comprising:

receiving, in a computing apparatus, a plurality of sets of itemized data, wherein each respective set of the itemized data includes a plurality of data fields containing an item description of a product purchased by a customer from a merchant;

performing, by the computing apparatus, natural language processing on item descriptions in the itemized data;

classifying, by the computing apparatus, items described by the item descriptions into item tiers based on the natural language processing, including;

determining presence of predetermined keywords in the item descriptions, and

scoring the item tiers of the items based on predetermined weights of the predetermined keywords; and

generating, by the computing apparatus, profiles of merchants and profiles of customers based at least in part on the item tiers classified from the natural language processing of the itemized data, wherein the merchants and the customers are identified in the itemized data.

2. The method of claim 1, wherein the natural language processing includes filtering from the item descriptions to remove words on a predetermined list.

3. The method of claim 1, wherein the natural language processing includes filtering from the item descriptions to remove predetermined patterns of words.

4. The method of claim 1, wherein each item tier is selected from a predetermined set of candidates.

5. The method of claim 1, wherein each item tier is a numerical value in a predetermined range.

6. The method of claim 1, wherein the natural language processing is based on a predictive model of item tier calculated from item description.

7. The method of claim 6, further comprising:

obtaining a training set of item descriptions and corresponding item tiers assigned by human representatives to item descriptions in the training set; and

training, by the computing apparatus, the predictive model using the training set of item descriptions and the corresponding item tiers.

8. The method of claim 1, further comprising:

extracting unit prices and unit counts from the itemized data; and

determining, from the unit prices and the unit counts, means prices of items described in the itemized data;

wherein the profiles of the merchants and the profiles of the customers are further based on the means prices of the items described in the itemized data.

9. The method of claim 8, further comprising:

determining, from the unit counts, item volumes of the merchants in the itemized data;

wherein the profiles of the merchants and the profiles of the customers are further based on the item volumes of the merchants in the itemized data.

10. The method of claim 9, further comprising:

determining a price tier for a first merchant from comparing a mean item price of merchants selling a product and a mean item price of the first merchant selling the product.

11. The method of claim 10, further comprising:

determining a volume tier for the first merchant from comparing a median item volume of merchants selling the product and an item volume of the first merchant selling the product.

12. The method of claim 11, wherein a profile of the first merchant is based on the price tier, the volume tier, and an item tier for the product sold by the first merchant.

13. The method of claim 11, further comprising:

combining volume tiers of products sold by the first merchant in a merchant category into a volume tier of the merchant category for the first merchant;

combining price tiers of the products sold by the first merchant in the merchant category into a price tier of the merchant category for the first merchant; and

combining item tiers of the products sold by the first merchant in the merchant category into an item tier of the merchant category for the first merchant.

14. The method of claim 13, wherein the combining of the volume tiers, the combining of the price tiers, and the combining of the item tiers are in a form of average weighted according to respective item volumes of the products sold by the first merchant in the merchant category.

15. The method of claim 9, further comprising:

classifying the merchants into merchant categories based on;

item tiers of items sold by the merchants;

mean item prices of the items sold by the merchants; and

item volumes of the items sold by the merchants.

16. The method of claim 15, further comprising:

classifying the customers into customer categories based on;

item tiers of items purchased by the customers;

mean item prices of the items purchased by the customers; and

item volumes of the items purchased by the customers.

17. The method of claim 15, further comprising:

classifying the customers into customer categories based on;

item tiers of items of merchants from which the customers made purchases;

mean item prices of the items of merchants from which the customers made purchases; and

item volumes of the items of merchants from which the customers made purchases.

18. The method of claim 16, further comprising:

matching the customer categories with the merchant categories; and

providing offers from merchants in a merchant category to customers in a customer category matching with the merchant category.

19. A computing apparatus, comprising:

a transaction handler configured to processing payment transactions in an electronic payment processing network;

a portal coupled with the transaction handler and configured to communicate with transaction terminals that initiate the payment transactions in the electronic payment processing network, wherein the portal is configured to receive, using a communication channel outside the electronic payment processing network, itemized data of items purchased via the payment transactions;

a data warehouse configured to store the itemized data and transaction data recording the payment transactions processed by the transaction handler;

a natural language processing engine configured to performing natural language processing on item descriptions in the itemized data, classify the items described by the item descriptions into item tiers based on the natural language processing by;

determining presence of predetermined keywords in the item descriptions, and

scoring the item tiers of the items based on predetermined weights of the predetermined keywords; and

a statistical analyzer configured to identify price tiers, volume tiers, wherein profiles of merchants and customers involved in the payment transactions are generated based at least in part on the item tiers, the price tiers, and the volume tiers.

20. A non-transitory computer storage medium storing instructions which when executed on a computing apparatus, cause the computing apparatus to perform a method, the method comprising:

receiving, in the computing apparatus, a plurality of sets of itemized data, wherein each respective set of the itemized data includes a plurality of data fields containing an item description of a product purchased by a customer from a merchant;

performing, by the computing apparatus, natural language processing on item descriptions in the itemized data;

classifying, by the computing apparatus, items described by the item descriptions into item tiers based on the natural language processing, including;

determining presence of predetermined keywords in the item descriptions, and

scoring the item tiers of the items based on predetermined weights of the predetermined keywords; and

generating, by the computing apparatus, profiles of merchants and profiles of customers based at least in part on the item tiers classified from the natural language processing of the itemized data, wherein the merchants and the customers are identified in the itemized data.