Systems and methods for matching data from an external catalog with data in an internal catalog

Patent metadata

Field	Value
Publication number	US-10776796-B2
Application number	US-201715605824-A
Country	US
Kind code	B2
Filing date	May 25, 2017
Priority date	May 25, 2017
Publication date	Sep 15, 2020
Grant date	Sep 15, 2020

Abstract

Systems and methods including one or more processing modules and one or more non-transitory storage modules storing computing instructions configured to run on the one or more processing modules and perform acts of receiving an external catalog comprising external attributes for each product of a plurality of products, mapping the external attributes from the external catalog to internal attributes for each product of the plurality of products in an internal catalog for an online retailer using an ensemble learning technique comprising a plurality of algorithms, incorporating the external attributes of the external catalog into the internal attributes in the internal catalog as mapped, and coordinating displaying of the external attributes and the internal attributes on a website of the online retailer.

First claim

What is claimed is:

1. A system comprising:
one or more processors; and
one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform:
receiving an external catalog from a first system that is external to the system, the external catalog comprising two or more respective external attributes for each respective product of a plurality of products; and
for each respective product of the plurality of products:
determining, for an internal catalog of an online retailer, a respective number of distinct values for each respective internal attribute of one or more internal attributes for each respective product of the plurality of products;
identifying, as a first group of the one or more internal attributes, the respective number of distinct values for at least one respective internal attribute of the one or more internal attributes that are above a catalog predetermined threshold;
in response to identifying the first group of the one or more internal attributes, mapping the two or more respective external attributes from the external catalog to one or more respective internal attributes that are not part of the first group of the one or more internal attributes for each respective product of the plurality of products using a first set of rules that comprise an ensemble learning technique comprising a plurality of algorithms, wherein:
a first respective external attribute of the two or more respective external attributes is mapped using the first set of rules that comprise the ensemble learning technique comprising the plurality of algorithms; and
a second respective external attribute of the two or more respective external attributes is mapped using a modified Jaro-Winkler algorithm utilizing a search pruning feature that compares bitsets of the second respective external attribute and the one or more respective internal attributes;

incorporating the two or more respective external attributes from the external catalog into the one or more respective internal attributes in the internal catalog, as mapped; and
coordinating displaying the two or more respective external attributes and the one or more respective internal attributes on a website of the online retailer.

2. The system of claim 1 , wherein the plurality of algorithms used in the ensemble learning technique of the first set of rules comprise two or more string distance algorithms.

3. The system of claim 2 , wherein the two or more string distance algorithms comprise at least one of a cosine similarity algorithm, a Levenshtein distance algorithm, a Jaro-Winkler distance algorithm, a n-gram distance algorithm, or a BM25 similarity algorithm.

4. The system of claim 1 , wherein mapping the two or more respective external attributes comprises, for each respective product of the plurality of products:
inputting the two or more respective external attributes and the one or more respective internal attributes into each algorithm of the plurality of algorithms;
determining the two or more respective external attributes match the one more respective internal attributes when a majority of respective outputs of the plurality of algorithms match the two or more respective external attributes with the one or more respective internal attributes; and
determining a respective confidence level that the two or more respective external attributes match the one or more respective internal attributes.

5. The system of claim 1 , wherein:
the first respective external attribute of the two or more respective external attributes for each respective product of the plurality of products comprises at least one respective external attribute name for each respective product of the plurality of products;
the second respective external attribute of the two or more respective external attributes for each respective product of the plurality of products comprises at least one respective external attribute value for each respective product of the plurality of products;
the one or more respective internal attributes for each respective product of the plurality of products comprise (1) at least one respective internal attribute name and (2) at least one respective internal attribute value for each respective product of the plurality of products; and
mapping the two or more respective external attributes comprises, for each respective product of the plurality of products:
mapping the at least one respective external attribute name to the at least one respective internal attribute name using the first set of rules that comprise the ensemble learning technique comprising the plurality of algorithms; and
mapping the at least one respective external attribute value to the at least one respective internal attribute value using the modified Jaro-Winkler algorithm utilizing the search pruning feature.

6. The system of claim 5 , wherein mapping the at least one respective external attribute value to the at least one respective internal attribute value using the modified Jaro-Winkler algorithm utilizing the search pruning feature comprises, for each respective product of the plurality of products:
determining a respective maximum possible score for the at least one respective external attribute value and the at least one respective internal attribute value; and
when the respective maximum possible score is greater than a first predetermined threshold, determining a Jaro-Winkler distance between the at least one respective external attribute value and the at least one respective internal attribute value using the modified Jaro-Winkler algorithm utilizing the search pruning feature.

7. The system of claim 6 , wherein determining the respective maximum possible score for the at least one respective external attribute value and the at least one respective internal attribute value comprises, for each respective product of the plurality of products:
creating a first respective bitset for the at least one respective external attribute value and a second respective bitset for the at least one respective internal attribute value; and
determining the respective maximum possible score by determining a respective maximum number of characters in common between the first respective bitset and the second respective bitset.

8. The system of claim 1 , wherein the one or more non-transitory storage devices storing the computing instructions are further configured to run on the one or more processors and perform:
determining, for each respective internal attribute of the one or more internal attributes of the plurality of products, at least one of (1) a respective average minimum number of characters, (2) a respective average maximum number of characters, or (3) a respective standard deviation of a number of characters;
determining, for at least one respective internal attribute of the one or more internal attributes, at least one of (1) the respective average minimum number of characters is outside of a third predetermined threshold, (2) the respective average maximum number of characters is outside of the third predetermined threshold, or (3) the respective standard deviation of the number of characters is outside of the third predetermined threshold; and
withholding the at least one respective internal attribute of the one or more internal attributes that is outside of the third predetermined threshold from, for each respective product of the plurality of products, mapping the two or more respective external attributes from the external catalog to the one or more respective internal attributes for each respective product of the plurality of products in the internal catalog.

9. The system of claim 1 , wherein:
the plurality of algorithms used in th

Assignees

• Wal Mart Stores Inc
• Walmart Apollo Llc

Inventors

• Amarthaluri Abilash
• Jain Chandni
• Nafar Zuzar Fakhruddin
• Deep Rohit

Classifications

• G06Q10/087

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title

• G06Q30/0201 Primary

  Market modelling; Market analysis; Collecting market data · CPC title