AUTOMATED CATEGORIZATION OF PRODUCTS IN A MERCHANT CATALOG

• US 20140172652A1
• Filed: 12/19/2012
• Published: 06/19/2014
• Est. Priority Date: 12/19/2012
• Status: Active Grant

• Current Assignee: Yahoo AD Tech LLC
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Pobbathi, Nageswara R., Dong, Anlei, Chang, Yi
• Granted Patent: US 10,528,907 B2
• Time in Patent Office: Days
• US Class Current: 705/28
• CPC Class Codes: G06Q 10/087 Inventory or stock management, e.g. order filling, procurement or balancing against orders

Abstract

A system and method is described for large-scale, automated classification of products. The system and method receives information about products, wherein such information includes one or more text metadata fields associated with each product, receives a set of categories, and automatically selects one or more categories from the set of categories to which each product belongs based upon at least one of the one or more text metadata fields associated with each product. A machine learning classifier may be used to automatically select the one or more categories to which each product belongs by operating upon a feature vector for each product derived from text metadata fields of the product description. The machine learning classifier may be trained using a set of pre-categorized product descriptions. The product-category associations generated by the system and method can be used to improve search engine results or product recommendations to consumers.

Claims

1. A method implemented on at least one machine having at least one processor, storage, and a communication platform connected to a network, comprising:

receiving information about a product, the information about the product comprising one or more text metadata fields;

receiving a set of categories comprising a plurality of classification trees, the plurality of classification trees storing the set of categories in a hierarchical manner; and

automatically selecting one or more categories from the set of categories to which the product belongs based upon at least one of the one or more text metadata fields, wherein the product belongs to more than one classification tree among the plurality of classification trees.

2. The method of claim 1, wherein automatically selecting the one or more categories from the set of categories to which the product belongs based upon at least one of the one or more text metadata fields comprises:

estimating a probability that the product belongs to each category in the set of categories based upon at least one of the one or more text metadata fields; and

associating the product with a particular category in the set of categories if the estimated probability that the product belongs to the particular category is among a set of highest estimated probability results for the set of categories.

3. The method of claim 2, wherein estimating the probability that the product belongs to each category in the set of categories based upon at least one of the one or more text metadata fields comprises:

generating a feature vector that represents the product based on at least one of the one or more text metadata fields; and

processing the feature vector in a machine learning classifier that estimates the probability that the product belongs to each category in the set of categories based upon the feature vector.

4. The method of claim 3, wherein generating the feature vector that represents the product based on at least one of the one or more text metadata fields comprises:

concatenating the one or more text metadata fields into a paragraph;

extracting one or more tokens from the paragraph; and

deriving a frequency number for each token by counting instances of each token in the paragraph.

5. The method of claim 4, wherein generating the feature vector that represents the product based on at least one of the one or more text metadata fields further comprises:

weighting the frequency number associated with each token based on which text metadata field the token came from.

6. The method of claim 3, wherein processing the feature vector in a machine learning classifier comprises:

processing the feature vector in a Naï

ve Bayes machine learning classifier.

7. The method of claim 3, wherein processing the feature vector in a machine learning classifier comprises:

processing the feature vector in a machine learning classifier that has been trained using at least one product catalog that includes text information about one or more products and categories associated therewith.

8. The method of claim 1, wherein the one or more text metadata fields include a title associated with the product, a description of the product, and a brand name associated with the product.

9. The method of claim 1, further comprising:

providing the selected one or more categories in association with the product to a search engine or a recommendation engine.

10. A system, comprising:

one or more computing machines;

a product classifier that is executed by at least one of the one or more computing machines to receive information about products, the information about the products comprising one or more text metadata fields associated with each of the products, to receive a set of categories, and to automatically associate each product with one or more categories from the set of categories based upon at least one of the one or more text metadata fields associated with the product; and

a search engine that is executed by at least one of the one or more computing machines to receive a search query, to determine that the search query is associated with a first product, to determine that the first product is associated with a particular category based at least upon the associations between categories and products generated by the product classifier, and to identify a second product associated with the particular category based at least upon the associations between categories and products generated by the product classifier.

11. The system of claim 10, wherein the search engine is executed by at least one of the one or more computing machines to determine that the search query is associated with a particular category and to identify at least one product associated with the particular category based at least upon the associations between categories and products generated by the product classifier.

13. The system of claim 10, wherein the search engine is executed by at least one of the one or more computing machines to determine that the search query is associated with a particular brand name, and to identify products associated with the brand name that are also associated with a plurality of different categories based at least upon the associations between categories and products generated by the product classifier.

14. The system of claim 10, wherein the product classifier is executed by at least one of the one or more computing machines to estimate a probability that a product belongs to each category in the set of categories based upon at least one of the one or more text metadata fields, and associate the product with a particular category in the set of categories if the estimated probability that the product belongs to the particular category is among a set of highest estimated probability results for the set of categories.

15. The system of claim 14, wherein the product classifier is executed by at least one of the one or more computing machines to generate a feature vector that represents the product based on at least one of the one or more text metadata fields, and process the feature vector in a machine learning classifier that estimates the probability that the product belongs to each category in the set of categories based upon the feature vector.

16. The system of claim 15, wherein the product classifier is executed by at least one of the one or more computing machines to generate the feature vector by concatenating the one or more text metadata fields into a paragraph, extracting one or more tokens from the paragraph, and deriving a frequency number for each token by counting instances of each token in the paragraph.

12. (canceled)

17. A system, comprising:

one or more computing machines;

a product classifier that is executed by at least one of the one or more computing machines to receive information about products, the information about the products comprising one or more text metadata fields associated with each of the products, to receive a set of categories based upon the one or more text metadata fields associated with the product; and

a recommendation engine that is executed by at least one of the one or more computing machines to determine that a user has accessed a document associated with a first product, to determine that the first product is associated with a particular category based at least upon the associations between categories and products generated by the product classifier, and to identify a second product associated with the particular category for recommendation to the user based at least upon the associations between categories and products generated by the product classifier.

18. The system of claim 17, wherein the product classifier is executed by at least one of the one or more computing machines to estimate a probability that a product belongs to each category in the set of categories based upon the one or more text metadata fields, and associate the product with a particular category in the set of categories if the estimated probability that the product belongs to the particular category is among a set of highest estimated probability results for the set of categories.

19. The system of claim 18, wherein the product classifier is executed by at least one of the one or more computing machines to generate a feature vector that represents the product based on the one or more text metadata fields, and process the feature vector in a machine learning classifier that estimates the probability that the product belongs to each category in the set of categories based upon the feature vector.

20. The system of claim 19, wherein the product classifier is executed by at least one of the one or more computing machines to generate the feature vector by concatenating the one or more text metadata fields into a paragraph, extracting one or more tokens from the paragraph, and deriving a frequency number for each token by counting instances of each token in the paragraph.