Systems and methods for automatic item classification

• US 10,120,929 B1
• Filed: 08/11/2014
• Issued: 11/06/2018
• Est. Priority Date: 12/22/2009
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Khoshnevisan, Cyrus, Kansal, Abhishek, Nguyen, Lam D., Touloumtzis, Michael Paul
• Primary Examiner(s): Perveen, Rehana
• Assistant Examiner(s): Wong, Huen
• Application Number:
  US14/457,019
• Time in Patent Office: 1,548 Days
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/901 Indexing; Data structures therefor; Storage structures for retrieval from the web G06F16/951

Abstract

An item categorization service is described that automatically categorizes items of interest to a user. The user may possess an item that they wish to offer for sale using a network-based service. The user may submit item information to the item categorization service to categorize the item of interest. Upon receipt, the categorization service may assess the relevance of the item information to hierarchically organized categories maintained by the network-based service. Categories having the highest relevance may be identified as first category candidates. The deepest common ancestor of the first category candidates may be identified the first category. One or more categories, representing sub-categories of the first category, may be identified as and subjected to relevance assessment. Those sub-categories having the highest relevance may be identified as second category candidates. The deepest common ancestor of the second category candidates may be identified as a second category for the item of interest.

Claims

1. A system for categorizing items, the system comprising:

a network-based service configured to;

receive a request to categorize an item of interest, the request including item information regarding the item of interest; and

transmit the request to an item categorization service;

a data store configured to store categories for items, wherein at least a portion of the categories are hierarchically organized with respect to one another; and

the item categorization service, wherein the item categorization service is in communication with the data store and the network-based service, and wherein the item categorization service is adapted to;

assign a first category to the item of interest by;

identifying a plurality of first category candidates based at least in part upon a relevance value associated with the item information and one or more of the categories for items stored by the data store, wherein at least a portion of the first category candidates are hierarchically organized;

determining which of the first category candidates has the most first category candidate attributes of the first category candidates; and

selecting the first category candidate that has the most first category candidate attributes as the first category of the item of interest; and

assign a second category to the item of interest by;

determining one or more rules associated with the first category of the item of interest;

identifying a plurality of second category candidates using the rules; and

selecting a second category of the item of interest from the plurality of second category candidates.

2. The system of claim 1, wherein the item categorization service is further configured to transmit updated item information, including the first and second categories of the item of interest, to the network-based service.

3. The system of claim 1, wherein item categorization service is further adapted to assign the second category to the item of interest based at least in part on a confidence level associated with the item information and the plurality of first category candidates.

4. The system of claim 1, wherein the first and second categories each comprise one or more category attributes and associated category attribute values.

5. The system of claim 4, wherein each of the second category candidates comprises a selected category attribute and associated category attribute value of the first category.

6. A system for categorizing items, the system comprising:

a data store adapted to store a plurality of hierarchically organized categories for items, each of the hierarchically organized categories including at least one category attribute and corresponding category attribute value; and

an item categorization service in communication with the data store, the item categorization service adapted to;

receive item information associated with an item of interest, the item information comprising at least one item attribute and corresponding item attribute value for the item of interest;

assign a first category for the item of interest from the hierarchically organized categories stored by the data store using the item information;

determine one or more rules associated with the assigned first category;

identify one or more second category candidates using the one or more rules and a relevance value associated with the item information and the at least one second category candidate;

determine which of the second category candidates has the most second category candidate attributes of the second category candidates; and

select the second category candidate that has the most second category candidate attributes as the second category of the item of interest.

7. The system of claim 6, wherein the item categorization service is further adapted to assign the first category for the item of interest based at least in part on relevance values associated with at least a portion of the hierarchically organized categories and the item information.

8. The system of claim 7, wherein the item categorization service is further adapted to assign the first category for the item of interest by identifying the hierarchically organized categories having the highest relevance values as first category candidates.

9. The system of claim 8, wherein the item categorization service is further adapted to assign the first category for the item of interest by selecting a deepest common ancestor of the first category candidates, wherein the deepest common ancestor comprises the first category candidate that is a common ancestor to more of the first category candidates than any other first category candidates.

10. The system of claim 6, wherein the item classification service is further adapted to identify a category attribute of the assigned first category and select the first category from a plurality of first category candidates, and wherein at least one of the first categories is not associated with the identified category attribute of the assigned first category.

11. The system of claim 6, wherein the item classification service is further adapted to identify a category attribute and corresponding category attribute value associated with the second category candidate and determine relevance values associated with the at least one category attribute and corresponding category attribute value of the at least one second category candidate and the received at least one item attribute and corresponding item attribute value.

12. The system of claim 6, wherein the item categorization service is further adapted to update the item information for the item of interest with the assigned first category and the assigned second category and to transmit the updated item information to the data store.

13. The system of claim 6, wherein the item classification service is further adapted to identify a category attribute of the assigned first category using the one or more rules.

14. The system of claim 6, wherein the item classification service is further adapted to assign the second category of the item of interest by selecting a deepest common ancestor of hierarchically organized selected second category candidates, wherein the deepest common ancestor comprises the selected second category candidate in the hierarchically organized selected second category candidates that is common to more of the hierarchically organized selected second category candidates than any other selected second category candidate.

15. A computer-implemented method for categorizing items, the method comprising:

under control of one or more computer systems configured with specific instructions;

receiving item information associated with an item of interest, the item information comprising one or more item attributes and corresponding item attribute values;

assigning a primary category describing the item of interest by selecting the primary category using a first search index, wherein the first search index comprises one or more hierarchically organized first categories having first attributes and corresponding first attribute values;

determining one or more rules associated with the assigned primary category;

obtaining a second search index comprising one or more hierarchically organized second categories using the one or more rules;

identifying one or more secondary category candidates from the second categories based at least in part upon relevance values associated with the item information and each of the one or more hierarchically organized second categories;

determining which of the second category candidates has the most second category candidate attributes of the second category candidates; and

selecting the second category candidate that has the most second category candidate attributes as a secondary category describing the item of interest.

16. The computer-implemented method of claim 15, wherein assigning the primary category describing the item of interest from the first search index further comprises determining relevance values associated with at least one of the first attributes and associated first attribute values of respective first categories and the received item attributes and item attribute values.

17. The computer-implemented method of claim 16, wherein assigning the primary category describing the item of interest from the first search index further comprises identifying a selected number of the first categories based on their relevance values as first category candidates.

18. The computer-implemented method of claim 17, wherein assigning the primary category describing the item of interest using the first search index further comprises selecting a deepest common ancestor of the first category candidates as the primary category, wherein the deepest common ancestor comprises the first category candidate that is common to more of the first category candidates than all other first category candidates.

19. The computer-implemented method of claim 15, wherein at least one of the first categories included in the first search index does not comprise a category attribute or the primary category.

20. The computer-implemented method of claim 15, further comprising determining relevance values associated with at least one category attribute and respective category attribute values of respective second categories and the received item attributes and item attribute values.

21. The computer-implemented method of claim 15, further comprising updating the item information with the assigned primary category and the assigned secondary category and transmitting the updated item information for storage.

22. A computer-readable medium having instructions stored thereon, wherein the instructions, when executed by a computing apparatus, cause the computing apparatus to:

receive item information associated with an item of interest, the item information comprising one or more item attributes and corresponding item attribute values;

assign a first category for the item of interest from hierarchically organized first categories stored by a data store;

determine one or more rules associated with the assigned first category;

identify one or more hierarchically organized second category candidates using the one or more rules and relevance values associated with the item information and each of the second category candidates;

determine which of the second category candidates has the most second category candidate attributes of the second category candidates; and

select the second category candidate that has the most second category candidate attributes as the second category of the item of interest.

23. The computer-readable medium of claim 22, wherein the instructions that cause the computing apparatus to assign the first category for the item of interest from the first categories comprise instructions that cause the computing apparatus to determine relevance values between at least one category attribute and category attribute value of respective first categories and the received item attributes and corresponding item attribute values.

24. The computer-readable medium of claim 23, wherein the instructions that cause the computing apparatus to assign the first category for the item of interest from the first categories further comprise instructions that cause the computing apparatus to identifying first category candidates from first categories using the determined relevance values.

25. The computer-readable medium of claim 24, wherein the instructions that cause the computing apparatus to assign the first category for the item of interest from the first categories further comprise instructions that cause the computing apparatus to select a deepest common ancestor of the first category candidates, wherein the deepest common ancestor comprises the first category candidate that is common to more first category candidates than any other first category candidate.

26. The computer-readable medium of claim 22, wherein at least one of the first categories does not comprise a category attribute of the assigned first category.

27. The computer-readable medium of claim 22, wherein the instructions further cause the computing apparatus to determine the relevance values between the item information and each of the second category candidates by determining relevance values between at least one second category candidate attribute and associated second category candidate attribute value and the received item attributes and item attribute values.

28. The computer-readable medium of claim 22, further comprising instructions that cause the computing apparatus to update the item information with the assigned first category and the assigned second category and to transmit the updated item information to a data store.