Retail recommendation domain model

• US 7,698,170 B1
• Filed: 08/05/2004
• Issued: 04/13/2010
• Est. Priority Date: 08/05/2004
• Status: Active Grant

• Current Assignee: Versata Development Group, Inc. (ESW Capital, LLC)
• Original Assignee: Versata Development Group, Inc. (ESW Capital, LLC)
• Inventors: Leffler, Jacob D., Dillon, Thomas H., Thier, Neil C., Darr, Timothy P., Singh, James, Packebush, Sherrill J., Wacksman, Jeremy S., Grant, Michael K.
• Primary Examiner(s): Airapetian; Mila
• Application Number:
  US10/912,743
• Time in Patent Office: 2,077 Days
• Field of Search: 705/26, 705/27
• US Class Current: 705/26.35
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0609 Buyer or seller confidence or verification

Abstract

A data processing system normalizes data sets (such as low-resolution transaction data) into high-resolution data sets by mapping generic information into attribute-based specific information that is stored in a database. By establishing a shared domain model for representing items in the recommendation context, catalog and quote history with common terms and concepts, a recommendation engine operating in the shared domain may process the attribute-based representations to make specific and relevant recommendations to the customer. In addition, when certain attribute values are normalized over time, recommendations derived from past order history can be intelligently applied to current orders. The normalized representation of elements in the shared domain may also be used to generate compelling selling point text for each recommendation that is specific to the marketing objectives of the seller and identifies the objectives of the buyer.

Claims

1. A computer-based method of generating recommendations for potential purchase by a customer, comprising:

transforming a part catalog into a normalized part catalog by mapping at least one generic catalog item from the part catalog into a set of normalized catalog items based on attribute names and values corresponding to the generic catalog item;

transforming a transaction history data set into a normalized transaction history data set with a computer system by mapping at least one generic item from the transaction history data set into a set of normalized items based on attribute names and values corresponding to the generic item;

generating association rules based on the normalized transaction history data set;

transforming a recommendation context into a normalized transaction data set by mapping at least one customer item from the recommendation context into a set of normalized transaction items based on attribute names and values corresponding to the customer item;

generating a recommendation by selecting a rule from the association rules that best matches with the normalized transaction data set.

2. The method of claim 1, where each of the transforming steps uses a domain model schema to transform the part catalog, transaction history data set and recommendation context into a shared domain, where the domain model schema comprises part group descriptors, feature descriptors, function descriptors and/or attribute descriptors describing each sellable part in the shared domain.

3. The method of claim 1, where the transforming of the transaction history data set into the normalized transaction history data set uses a domain model schema comprising a set of attributes describing specific aspects of sellable parts in a shared domain, where the set of attributes are selected to reduce a dimensionality measure of the generated association rules.

4. The method of claim 1, where one or more of the transforming steps converts any SKU descriptor into at least one normalized attribute-based descriptor.

5. The method of claim 4, where the attribute-based descriptor comprises a qualitative descriptor.

6. The method of claim 4, where the attribute-based descriptor comprises a time-variant descriptor.

7. The method of claim 1, where a new item in the part catalog, transaction history data set or recommendation context that corresponds to an existing part group is translated to at least one attribute-based descriptor for the existing part group.

8. The method of claim 7, comprising comparing a part group for the new item to an existing part group.

9. The method of claim 1, comprising generating a recommendation message to accompany the recommendation by using the normalized transaction data set to select a selling point text based on at least one attribute contained in a selected association rule.

10. The method of claim 9, comprising:

selecting a profile estimate from a plurality of profiles in response to the normalized transaction data set, where the profile estimate comprises a profile lead-in text portion and an attribute selling point text portion; and

constructing the recommendation message from the profile lead-in text portion and the attribute selling point text portion.

11. The method of claim 1, where the generation of a recommendation comprises:

scoring at least one recommendation message based upon how closely a profile associated with the recommendation message corresponds to the recommendation context;

scoring at least one association rule based upon how closely the association rule corresponds to the recommendation context;

selecting a recommendation message for inclusion with the recommendation based upon said scoring steps.

12. The method of claim 1, comprising generating a recommendation message by using a recommendation context to select a profile from a plurality of user profiles identifying use, goals, and/or objectives of a customer associated with the recommendation context, said selected profile comprising a recommendation message.

13. The method of claim 12, wherein the selected profile comprises a set of attributes representing attributes that a user of selected profile type would be interested in when making a purchase.

14. The method of claim 12, wherein the selection of the profile from the plurality of user profiles uses the domain model schema to score each user profile based upon how closely the user profile corresponds to the recommendation context.

15. The method of claim 12, wherein two or more profiles are selected from the plurality of user profiles by using at least one normalized transaction item to score each user profile based upon how closely each user profile corresponds to the recommendation context, thereby selecting at least two user profiles corresponding most closely to the recommendation context as said two or more profiles.

16. The method of claim 1, comprising generating a recommendation message that comprises a selling point text for an attribute or attribute value of a recommended item that explains why the recommendation should be accepted.

17. The method of claim 1, wherein the normalized part catalog is used to identify a specific recommended item in response to the generated recommendation.

18. The method of claim 1, wherein the transforming of the part catalog into a normalized part catalog comprises using a web page catalog scraper to extract attribute values, part descriptions and/or other part data from a web page.

19. The method of claim 1, wherein the transforming of the part catalog into a normalized part catalog comprises using a schema translator to transform a native catalog schema to the domain model schema.

20. A system for generating purchase recommendations for a customer, comprising:

a database for storing a transaction history data set;

means for transforming the transaction history data set into an attribute-based transaction history data set;

means for generating association rules from the attribute-based transaction history data set;

means for transforming a recommendation context generated by a customer into an attribute-based transaction data set; and

a recommendation engine for generating a recommendation by selecting a rule from the association rules that best matches with the attribute-based transaction data set.