Searching product catalogs

• US 20020138481A1
• Filed: 03/23/2001
• Published: 09/26/2002
• Est. Priority Date: 03/23/2001
• Status: Active Grant

• Current Assignee: PayPal, Inc. (PayPal Holdings, Inc.)
• Original Assignee: International Business Machines Corporation
• Inventors: Aggarwal, Gaurav, Ghosal, Sugata
• Granted Patent: US 6,728,706 B2
• Time in Patent Office: Days
• US Class Current: 707/6
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

Searching online catalog databases by shoppers is improved by performing similarity searching on searches performed by the shopper, in conjunction with adjusting the similarity metric used during the search to interactively improve the relevance of the resulting search results to the shopper. This involves using relevance feedback and/or product redefinition to learn the “implied concept” of the shopper'"'"'s stated product requirements; that is, the quality of the search results is enhanced through understanding the concept implied by the shopper'"'"'s queries by, for example, learning from the product results marked as “relevant” or “irrelevant” by the shopper. This approach is in contrast with attempts to enhance search results using, for example, the shopper'"'"'s past orders or previous actions, or the past orders or previous actions of other shoppers who may have a purchase history similar to that of the current shopper.

Claims

1. A method of providing to a user an interactive product search facility suitable for searching product information stored in a product catalog, the method comprising the steps of:

(a) providing a database of records respectively describing products, each of said records including a plurality of fields containing respective field values which characterise said products;

(b) receiving from a user a request to search said database of records in the form of a query based on one or more of said fields;

(c) providing to the user an initial set of similar search records found as a result of a similarity search based on said search request;

(d) receiving from the user relevance feedback as to which of said similar search records are of greater and/or lesser relevance to the user compared with other similar search records; and

(e) providing to the user a modified set of similar search results found as a result of a similarity search based on (i) said search request, and (ii) said relevance feedback, said modified set of similar search results providing to the user records which are generally of greater relevance to the user than said initial set of similar search results.

2. The method as claimed in claim 1, wherein the provision of similar search records involves a combination of two or more different types of product attributes, including numeric, nominal and image-based product attributes.

3. The method as claimed in claim 1, wherein the provision of similar search results is performed with the assistance of similarity functions.

4. The method as claimed in claim 3, wherein different similarity functions are used for different types of product attributes.

5. The method as claimed in claim 4, wherein the similarity function used for numeric product attributes is Mahalanobis distance.

6. The method as claimed in claim 5, wherein the Mahalanobis distance is a weighted Euclidean distance measure.

7. The method as claimed in claim 6, wherein to calculate the weighted Euclidean distance measure, relative weights are assigned to respective product attributes to represent the relative importance of that attribute in the similarity function.

8. The method as claimed in claim 1, wherein the relevance feedback involves an indication from the user, for one or more of the results in the set of similar search results, at least whether the result is relevant or irrelevant.

9. The method as claimed in claim 1, wherein on the basis of the relevance feedback, the similarity function, or parameters thereof, are modified so that the results of the modified similarity search provide the modified set of similar search results which better matches those of the initial set of similar search results designated relevant or irrelevant.

10. The method as claimed in claim 7, wherein the relevance feedback received from the user is used to automatically adjust the relative weights assigned to respective product attributes.

11. The method as claimed in claim 1, wherein steps (c) and (d) are iteratively performed a plurality of times.

12. The method as claimed in claim 1, wherein the initial set of similar search results includes K-nearest neighbours found as a result of the similarity search.

13. The method as claimed in claim 1, wherein the initial set of similar search results includes search results that are markedly different from each other and the query point, and are selected to assist in orienting the similarity function to the shopper'"'"'s implied concept.

14. The method as claimed in claim 1, wherein before providing an initial set of similar search results, the following steps are performed:

(i) generating a number of similar product specifications by modifying one product attribute for each such product specification; and

(ii) receiving relevance feedback for one or more of these modified product specification, in order to adjust the query point and learn the similarity metric that is most suitable, prior to conducting a similarity search on the basis of search request.

15. A computer program product able to access a computer readable medium having recorded therein a set of instructions comprising a computer program for searching product catalogs, the computer program comprising:

(a) code means for providing a database of records respectively describing products, each of said records including a plurality of fields containing respective field values which characterise said products;

(b) code means for receiving from a user a request to search said database of records in the form of a query based on one or more of said fields;

(c) code means for providing to the user an initial set of similar search records found as a result of a similarity search based on said search request;

(d) code means for receiving from the user relevance feedback as to which of said similar search records are of greater and/or lesser relevance to the user compared with other similar search records; and

(e) code means for providing to the user a modified set of similar search results found as a result of a similarity search based on (i) said search request, and (ii) said relevance feedback, said modified set of similar search results providing to the user records which are generally of greater relevance to the user than said initial set of similar search results.

16. An apparatus searching product catalogs, the apparatus comprising:

(a) means for providing a database of records respectively describing products, each of said records including a plurality of fields containing respective field values which characterise said products;

(b) means for receiving from a user a request to search said database of records in the form of a query based on one or more of said fields;

(c) means for providing to the user an initial set of similar search records found as a result of a similarity search based on said search request;

(d) means for receiving from the user relevance feedback as to which of said similar search records are of greater and/or lesser relevance to the user compared with other similar search records; and

(e) means for providing to the user a modified set of similar search results found as a result of a similarity search based on (i) said search request, and (ii) said relevance feedback, wherein said modified set of similar search results provides to the user records which are generally of greater relevance to the user than said initial set of similar search results.