Weighted preference inference system and method

• US 7,836,057 B1
• Filed: 05/23/2007
• Issued: 11/16/2010
• Est. Priority Date: 09/24/2001
• Status: Expired due to Term

• Current Assignee: Viteds LLC
• Original Assignee: Auguri Corporation (Element Data, Inc.)
• Inventors: Yang, Jie, Micaelian, Fadi Victor, Criswell, David Brandon, Sawey, Richard, Scoffone, Emil Mario
• Primary Examiner(s): Rones; Charles
• Assistant Examiner(s): Khoshnoodi; Fariborz
• Application Number:
  US11/752,605
• Time in Patent Office: 1,273 Days
• Field of Search: 707/3, 707/4, 707/104.1, 707/2
• US Class Current: 707/748
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

A search engine for databases, data streams, and other data sources allows user preferences as to the relative importance of search criteria to be used to rank the output of the search engine. A weighted preference generator generates weighted preference information including at least a plurality of weights corresponding to a plurality of search criteria. A weighted preference data search engine uses the weighted preference information to search a data source and to provide an ordered result list based upon the weighted preference information. A method for weighted preference data searching further includes allowing a user to re-rank the ordered result list and determine newly updated weighted preferences based on the re-ranking process. A plurality of weights signifying the relative importance of the search criteria may also be generated by the user inputting an ordered list. This ordered list may be a single alternative or a plurality of alternatives.

Claims

1. A computer implemented method for reconstructing a first search query corresponding to a user, and inferring user preferences manifested as query parameters, said computer implemented method comprising:

receiving a plurality of search results, said plurality of search results equivalent to a first output derived from executing said first search query;

determining a plurality of query parameters based on said plurality of search results, wherein said plurality of query parameters corresponds to user preferences, said plurality of search results includes an ordered results list, items from said ordered results list each have values for one or more specific criteria, and said values are associated with said items of said ordered results list;

generating, based on said ordered search results list, specific parameter values for said plurality of query parameters;

constructing a second search query based on said plurality of search results, said plurality of query parameters, and said specific parameter values;

wherein said second search query is equivalent to and a reconstruction of said first search query in that a second output derived from executing said second search query results in said plurality of search results,

whereby said first search query is reconstructed and said user preferences are inferred via said plurality of search results without further knowledge of said first search query.

2. The computer implemented method for reconstructing a first search query as recited in claim 1, wherein said ordered results list is obtained from a first database.

3. The computer implemented method for reconstructing a first search query as recited in claim 1, wherein said ordered results list is obtained from said user.

4. The computer implemented method for reconstructing a first search query as recited in claim 3, wherein said ordered results list obtained from said user is a list of items ranked in an order of desirability by said user.

5. The computer implemented method for reconstructing a first search query as recited in claim 4, wherein items in said list of items are automobiles.

6. The computer implemented method for reconstructing a first search query as recited in claim 1, further comprising selecting an algorithm for generating said specific parameter values based on a number of items in said ordered results list.

7. The computer implemented method for reconstructing a first search query as recited in claim 1, further comprising allowing said user to adjust said user preferences and generating a third search query in response to said user adjusting said user preferences.

8. The computer implemented method for reconstructing a first search query as recited in claim 1, further comprising displaying said inferred user preference to said user in a graphical format;

allowing said user to adjust said inferred user preferences through a graphical user interface; and

generating a third search query in response to said user adjusting said inferred user preferences.

9. A computer implemented method for constructing a search query based on inferred user preferences for a specific user, said search query including a plurality of query parameters having specific values indicative of user preferences, said computer implemented method comprising:

receiving an ordered list of items selected from a first database and ordered by said specific user, each of said items having item specific values for a plurality of criteria, said item specific values defined by said items without reference to user preferences, said ordered list of items each having values for one or more specific criteria, and said values are associated with said items of said ordered list of items; and

constructing said search query using said ordered list and said item specific values for said plurality of criteria, said constructing including;

capturing said ordered result list;

capturing said plurality of query parameters;

measuring values of said criteria for each item;

deriving said query parameters including importance weighting for each of the criteria that generates said ordered result list; and

wherein executing said constructed search query on said first database results in said ordered list of items.

10. A computer system for constructing a search query based on inferred user preferences for a specific user, said search query including a plurality of query parameters having specific values indicative of user preferences, said computer system comprising:

a receiver unit receiving an ordered list of items selected from a first database and ordered by said specific user, each of said items having item specific values for a plurality of criteria, said item specific values defined by said items without reference to user preferences, said ordered list of items each having values for one or more specific criteria, and said values are associated with said items of said ordered list of items; and

a search construction engine for constructing said search query using said ordered list and said item specific values for said plurality of criteria, said search construction engine further including;

a first engine that captures said ordered result list, said ordered result list of items having values for one or more said associated criteria values, and said values are associated with said items of said ordered result list of items;

a second engine that captures a plurality of query parameters;

a third engine that measures values of every criteria for each result item; and

a fourth engine that derives said query parameters including importance weighting for each of the criteria that generate said ordered result list.

11. A computer system for reconstructing a first search query corresponding to a user, and inferring user preferences manifested as query parameters, said computer implemented method comprising:

means for receiving a plurality of search results, said plurality of search results equivalent to a first output derived from executing said first search query;

means for determining a plurality of query parameters based on said plurality of search results, wherein said plurality of query parameters corresponds to user preferences, said plurality of search results includes an ordered results list, items from said ordered results list each have values for one or more specific criteria, and said inherent values are associated with said items of said ordered results list;

means for generating, based on said ordered search results list, specific parameter values for said plurality of query parameters;

means for constructing a second search query based on said plurality of search results, said plurality of query parameters, and said specific parameter values;

wherein said second search query is equivalent to and a reconstruction of said first search query in that a second output derived from executing said second search query results in said plurality of search results,

whereby said first search query is reconstructed and said user preferences are inferred via said plurality of search results without further knowledge of said first search query.

12. A method comprising:

receiving a first plurality of search items, wherein each item of the first plurality of search items has a corresponding plurality of criterion, and wherein each criterion of the plurality of criterion has a corresponding criterion value; and

computing a plurality of weighted parameters, wherein each weighted parameter of the plurality of weighted parameters corresponds to a respective criterion of the plurality of criterion and is weighted to indicate a preference for the respective criterion, and wherein a plurality of scores, each of the plurality of scores corresponding to respective items of the first plurality of items, calculated by applying some function to the weighted parameter and the respective criterion value of the corresponding criterion, results in items of the first plurality of items having the highest score;

wherein computing a plurality of weighted parameters comprises;

(i) creating a desirability vector p, wherein each element p_j of the desirability vector p indicates the desirability of a corresponding item relative to the other items of the first plurality of items;

(ii) calculating a plurality of criterion scores a_ji, wherein each criterion score of the plurality of criterion scores corresponds to the respective criterion, and wherein each criterion score of the plurality of criterion scores corresponds to a different item;

(iii) creating a scores matrix A from the plurality of criterion scores; and

(iv) multiplying the desirability vector p by the scores matrix A to yield a plurality of weighted parameters; and

wherein computing a weighted parameter w_i of the plurality of weighted parameters corresponding to the respective criterion comprises;

(v) creating a desirability vector p, wherein each element p_j of the desirability vector p indicates the desirability of a corresponding item relative to the other items of the first plurality of items;

(vi) calculating a plurality of criterion scores a_ji, wherein each criterion score of the plurality of criterion scores corresponds to the respective criterion and wherein each criterion score of the plurality of criterion scores corresponds to a different item; and

(vii) calculating a weighted parameter corresponding to the respective criterion by applying a scoring function ƒ

(w, a_ji) so that for every p_j, p_j=ƒ

(w, a_ji).

13. The method of claim 12 further comprising:

receiving user feedback as a second plurality of search items, wherein the second plurality of search items include the same items as the first plurality of items but can be ordered differently than the first plurality of items; and

computing a plurality of weighted parameters, wherein each weighted parameter of the plurality of weighted parameters corresponds to a respective criterion of the plurality of criterion and is weighted to indicate a preference for the respective criterion, and wherein a plurality of scores, each of the plurality of scores corresponding to respective items of the first plurality of ranked items, calculated by summing the products of each weighted parameter and the respective criterion value of the corresponding criterion results in the highest ordered item of the second plurality of items having the highest score.