Rule-based item classification

Patent metadata

Field	Value
Publication number	US-9483741-B2
Application number	US-201414473351-A
Country	US
Kind code	B2
Filing date	Aug 29, 2014
Priority date	Mar 28, 2013
Publication date	Nov 1, 2016
Grant date	Nov 1, 2016

Abstract

Systems and methods are disclosed herein for rule-based item classification. The methods include receiving, by a computing device, an item record for analysis. The computing device may determine ranked lists of item types using rule-based classifiers and machine learning-based classifiers. Then, the computing device may aggregate the ranked lists of item types to generate a combined ranked list of item types.

First claim

The invention claimed is:

1. A computer-implemented method for product classification, the method comprising:
receiving, by one or more processors of a server, a product record associated with a product, the product record comprising a description of the product;
generating, by the one or more processors, a first ranked list of product types based on the product record using a plurality of rule-based classifiers;
generating, by the one or more processors, a second ranked list of product types based on the product record using a plurality of machine learning-based classifiers;
aggregating, by the one or more processors, the first ranked list of product types and the second ranked list of product types to generate a combined ranked list of product types;
providing the description of the product and a portion of the combined ranked list of product types to a crowdsourcing computer network for evaluation; and
returning, from the crowdsourcing computer network, a validation decision, wherein:
when the crowdsourcing computer network determines that a type of the product is correct, the validation decision indicates that the type of the product is correct;
when the crowdsourcing computer network determines that the type of the product is not correct, the validation decision indicates that the type of the product is not correct; and
when the crowdsourcing computer network determines that the description of the product is insufficient to accurately judge whether the type of the product is correct, the validation decision indicates that the type of the product is unclear.

2. The method of claim 1 , wherein the generating, by the one or more processors, the first ranked list of product types based on the product record using the plurality of rule-based classifiers comprises: applying a whitelist rule and a blacklist rule to the product record to obtain the first ranked list of product types, the whitelist rule defining that the product belongs to a first product type, the blacklist rule defining that the product does not belong to a second product type different from the first product type.

3. The method of claim 1 wherein at least a portion of the combined ranked list of product types comprises a determined product type of the product; and wherein the method further comprises:
determining, based on the evaluation, that the determined product type is not acceptable; and
causing generation of a new rule and an updated product type by an analyst.

4. The method of claim 3 , further comprising:
updating at least one classifier of the plurality of rule-based classifiers using the new rule; and
training at least one classifier of the plurality of machine learning-based classifiers using the updated product type and the description of the product.

5. The method of claim 1 , further comprising: determining a third ranked list of product types based on the product record using a plurality of attribute-based classifiers.

6. The method of claim 5 , wherein the aggregating, by the one or more processors, the first ranked list of product types and the second ranked list of product types into the combined ranked list of product types comprises: aggregating the first ranked list of product types, the second ranked list of product types, and the third ranked list of product types into the combined ranked list of product types.

7. The method of claim 1 , further comprising: initializing at least a portion of the plurality of rule-based classifiers and the plurality of machine learning-based classifiers using a set of training product records and a basic rule.

8. The method of claim 1 , further comprising:
analyzing the product record to determine an initial product type based on an initial classification rule;
in response to a determination that: a confidence score associated with the initial product type and the product is greater than a predetermined threshold, providing the product record to the plurality of rule-based classifiers;
in response to a determination that: a confidence score associated with the initial product type and the product is greater than a predetermined threshold, providing the product record to the plurality of machine learning-based classifiers;
in response to a determination that: the initial classification rule is not related to the product record, providing the product record to the plurality of rule-based classifiers; and
in response to a determination that: the initial classification rule is not related to the product record, providing the product record to the plurality of machine learning-based classifiers.

9. The method of claim 1 , wherein the plurality of machine learning-based classifiers comprises at least one of a naive Bayes classifier, a k-Nearest Neighbors (KNN) classifier, or a Perceptron classifier.

10. The method of claim 1 comprising filtering the combined ranked list of product types to generate a filtered list using a blacklist rule.

11. One or more non-transitory computer-readable media storing computer-executable instructions that, when executed on one or more processors, cause the one or more processors to perform acts comprising:
receiving a product record associated with a product, the product record comprising a description of the product;
classifying the product based on the product record using a plurality of rule-based classifiers to generate a first ranked list of product types, including applying a whitelist rule and a blacklist rule to the product record to obtain the first ranked list of product types, wherein the whitelist rule defines that the product belongs to a first product type and the blacklist rule defines that the product does not belong to a second product type different from the first product type;
classifying the product based on the product record using a plurality of attribute-based classifiers to generate a second ranked list of product types;
combining the first ranked list of product types and the second ranked list of product types to generate a combined ranked list of product types;
providing at least a portion of the combined ranked list of product types to a crowdsourcing computer network for evaluation; and
returning, from the crowdsourcing computer network, a validation decision, wherein:
when the crowdsourcing computer network determines that a type of the product is correct, the validation decision indicates that the type of the product is correct;
when the crowdsourcinq computer network determines that the type of the product is not correct, the validation decision indicates that the type of the product is not correct; and
when the crowdsourcing computer network determines that the description of the product is insufficient to accurately judge whether the type of the product is correct, the validation decision indicates that the type of the product is unclear.

12. The one or more non-transitory computer-readable media of claim 11 , wherein the classifying of the product based on the product record using a plurality of rule-based classifiers to generate the first ranked list of product types comprises:
applying an attribute value-based classification rule.

13. The one or more non-transitory computer-readable media of claim 11 , wherein:
the at least the portion of the combined ranked list of product types comprises a determined product type of the product; and
when the validation decision is that the determined product type is not correct,
the acts further comprise causing generation of a new rule and an updated product type by an analyst.

14. The one or more non-transitory computer-readable media of claim 13 , wherein the acts further co

Assignees

• Wal Mart Stores Inc

Inventors

• Sun Chong
• Yang Fan
• Rampalli Narasimhan
• Lamba Digvijay Singh
• Xie Jun
• Chivers Thomas E
• Kavaturi Gokul
• Phung Tracy Thutrang

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• G06F18/254

  of classification results, e.g. of results related to same input data · CPC title

• G06F18/24765

  Rule-based classification · CPC title

• G06Q50/01

  Physics · mapped topic

• G06N99/005 Primary

  Physics · mapped topic