System and method of predicting subjective reactions

• US 4,870,579 A
• Filed: 10/01/1987
• Issued: 09/26/1989
• Est. Priority Date: 10/01/1987
• Status: Expired due to Term

• Current Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Original Assignee: Neonics, Inc.
• Inventors: Hey, John B.
• Primary Examiner(s): Lall, Parshotam S.
• Assistant Examiner(s): MELNICK, STEVEN A
• Application Number:
  US07/103,848
• Time in Patent Office: 726 Days
• Field of Search: 364/419, 364/401, 364/550, 364/554, 364/410, 434/237, 434/236, 358/86, 358/84, 455/2, 273/161
• US Class Current: 705/7.31
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G09B 5/00 Electrically-operated educational appliances working with questions and answers G09B7/00; simulators G09B9/00; advertising or displaying in general G09F
• CPC Class Codes: H04N 21/466 Learning process for intelligent management, e.g. learning user preferences for recommending movies services using the results of monitoring in broadcast systems H04H60/61

Abstract

A system and method of predicting, for a user selected from a group of users, the reactions of the selected user to items sampled by one or more users in the group but not sampled by the selected user. The predictions are based on other items previously sampled by that user. A scalar rating is defined for each item sampled by the selected user to represent the reaction of the selected user to that item. The selected user is successively paired with other users in the group who have defined scalar ratings for at least some of the items sampled by the selected user to determine the overall difference in ratings for items sampled by both members of each successive pair. One or more of the other users are designated as predicting users and a weighting value is assigned to each of the predicting users based on the overall difference in ratings between that predicting user and the selected user. The weighting values are applied for each item unsampled by the selected user to proportionally alter the difference between a rating previously predicted for each identified item and any actual ratings of that item by the predicting users to adjust the reaction predictions for the selected user.

Claims

1. A method of automatically predicting, for a user selected from a group of users, the reactions of the selected user to items sampled by one or more users in the group but not sampled by the selected user, the reaction predictions being based on other items previously sampled by that of the user, comprising:

defining, for each item sampled by the selected user, a scalar rating representing the reaction of the selected user to that item;

successively pairing the selected user with other users in the group for whom have been defined scalar timings for at least some of the items sampled by the selected user to determine the difference in ratings for items sampled by both members of each successive pair;

designating at least one of the other users as a predicting user and assigning a weighting value to each of the predicting users based on the difference in ratings between that predicting user and the selected user; and

applying the weighting values to items not yet sampled by the selected user to proportionally alter the difference between a rating previously predicted for each item not yet sampled by the selected user and the ratings of that item by the predicting users to adjust the reaction predictions for the selected user to more closely predict the actual reaction of the user to that item.

2. A method using a computing device to automatically predict, for a user a selected from a group of users, the reactions of the selected user to items sampled by one or more users in the group but not sampled by the selected user, the reaction predictions being based on other items previously sampled by that user, comprising:

defining, for each item sampled by the selected user, a scalar rating representing the reaction of the selected user to that item, the step of defining including entering, into an input device of the computing device, information representing the reaction of the selected user to items sampled by that user;

successively pairing the selected use with other users in the group for whom have been defined scalar ratings for at least some of the items sampled by the selected user to determine the difference in ratings for items sampled by both members of each successive pair;

designating at least one of the other users as a predicting user and as singing a weighting value to each of the predicting users based on the difference in ratings between that predicting user and the selected user; and

applying the weighting values to items not yet sampled by the selected user to proportionally alter the difference between a rating previously predicted for each item not yet sampled by the selected user and the ratings of that item by the predicting users to adjust the reaction predictions for the selected user to more closely predict the actual reaction of the user to that item.

3. The method of claim 2 in which pairing includes successively matching, for each pair, items sampled by both members and, for each matched item, comparing the ratings of one member from the rating of the other member to obtain the difference in ratings.

4. The method of claim 3 in which pairing further includes converting, for each pair, the difference in ratings for each matched item to a closeness value, and combining the closeness values for the members of that pair.

5. The method of claim 4 in which pairing further includes weighting, for each pair, the combined closeness values by the number of items sampled by both members to generate the agreement scalar.

6. The method of claim 2 in which designating and converting includes defining a greater weighting value for predicting users having a larger agreement scalar and defining a lesser weighting value for predicting users having a smaller agreement scalar.

7. The method of claim 2 in which designating and converting includes ranking the predicting users by ascending order of agreement scalar and in that order assigning to the ranked predicting users progressively larger weighting values.

8. The method of claim 2 in which applying includes combining the weighting value for each predicting user with the difference between the actual rating by that predicting user and the previous predicted rating by the selected user for each unsampled item, and summing the combination with the previously predicted rating.

9. The method of claim 2 in which pairing includes successively pairing the selected user with each other user in the group.

10. The method of claim 2 in which pairing includes defining a subset of other users in the group to be successively paired with the selected user.

11. The method of claim 2 further including successively selecting the remainder of the users in the group to adjust the reaction predictions for each user in the group.

12. A method of selectively recommending, for each user successively selected from a group of users, items sampled by one or more users in the group but not yet sampled by the selected user, the recommendations being based on other items previously sampled by that user and being represented by a scalar rating for each item, the method comprising:

defining, for each item sampled by the selected user, a scalar rating representing the reaction of the selected user to that item;

successively pairing the selected user with other users in the group for whom have been defined scalar ratings for at least some of the items sampled by the selected user to determine the difference in ratings for items sampled by both members of each successive pair;

generating for each pair an agreement scalar representing the overall rating agreement between the members of that pair;

designating at least one of the other users as recommending users;

converting, for each of the recommending users, the associated agreement scalar into a recommendation-fraction;

identifying items not yet sampled by the selected user;

establishing an initial scalar rating for each identified item for the selected user;

applying the recommendation-fractions to proportionally decrease the difference between one of the initial scalar rating for each identified item and a scalar rating previously established for each identified item for the selected user and the actual ratings of that item by the recommending users to adjust the recommendations for the selected user;

successively selecting the remainder of the users in the group to adjust the recommendations for each user in the group by the above-recited steps; and

presenting, for each user, a plurality of items based on the recommendations for that user.

13. The method of claim 12 in which pairing includes successively matching, for each pair, items sampled by both members and, for each matched item, comparing the rating of one member with the rating of the other member to obtain the difference in ratings.

14. The method of claim 13 in which pairing further includes converting, for each pair, the difference in ratings for each matched item to a closeness value, and combining the closeness values for the members of that pair.

15. The method of claim 14 in which pairing further includes weighting, for each pair, the combined closeness values by the number of items sampled by both members to generate the agreement scalar.

16. The method of claim 15 in which designating and converting includes sorting the recommending users by ascending order of agreement scalar and in that order assigning to the ranked predicting users progressively larger weighting values.

17. The method of claim 16 in which applying includes combining the weighting fraction for each recommending user with the difference between the actual rating by that recommending user and the previous predicted rating by the selected user for each identified item, and summing the combination with the previously established rating.

18. The method of claim 17 further including initially setting the established rating for each unsampled item for each selected user to a low rating value.

19. The method of claim 12 in which presenting includes providing a listing of a limited number of highly recommended items not yet sampled by the selected user.

20. The method of claim 12 in which establishing includes providing a low scalar rating value as the initial scalar rating.

21. A system for predicting, for a user selected from a group of users, the reactions of the selected user to items sampled by one or more users in the group but not sampled by the selected user, the predictions being based on other items previously sampled by that user, comprising:

means for defining, for each item sampled by the selected user, a scalar rating representing the reaction of the selected user to that item, said means for defining including input means for entering information representing the reaction of the selected user to items sampled by that user;

means for successively pairing the selected user with other users in the group for whom have been defined scalar ratings for at least some of the items sampled by the selected user to determine the difference in ratings for items sampled by both members of each successive pair;

means for designating at least one of the other users as a predicting user and assigning a weighting value to each of the predicting users based on the overall difference in ratings between that predicting user and the selected user; and

means for applying the weighting values to items not yet sampled by the selected user to proportionally alter the difference between a rating previously predicted for each item not yet sampled by the selected user and the ratings of that item by the predicting users to adjust the reaction predictions for the selected user to more closely predict the actual reaction of the user to that item.

22. A computing device for predicting, for a user selected from a group of users, the reactions of the selected user to items sampled by one or more users in the group but not sampled by that user, the prediction being based on other items previously sampled by that user, comprising:

means for defining, for each item sampled by the selected user, a scalar rating representing the reaction of the selected user to that item, said means for defining including input means for entering information representing the reaction of the selected user to items sampled by that user;

means for successively pairing the selected user with others users in the group for whom have been defined scalar ratings for at least some of the items sampled by the selected user to determine the difference in rating for items sampled by both members of each successive pair, said means for pairing including means for generating for each pair an agreement scalar representing the overall rating difference between the members of that pair;

means for designating at least one of the other users as a predicting user and for converting, for each of the predicting users, the agreement scalar into a weighing fraction;

means for establishing an initial scalar rating for each identified item for the selected user; and

means for identifying items not yet sampled by the selected user and applying the weighting values to items not yet sampled by the selected user to proportionally alter the difference between one of the initial scalar rating for each identified item and a rating previously predicted for each identified item and the ratings of that item by the predicting users to adjust the reaction predictions for the selected user to more closely predict the actual reaction of the user to that item.

23. The computing device of claim 22 in which said means for establishing establishes said initial scalar rating based on information entered through said input means.

24. The computing device of claim 22 in which said means for establishing includes memory for storing said initial scalar rating.

25. A method of using a computing device to automatically update, for a user selected from a group of users, estimated scalar ratings for items sampled by one or more users in the group but not yet sampled by the selected user to more closely predict the reaction of the selected user to the not-yet-sampled items, the method comprising:

defining, for each item sampled by the selected user, an actual scalar rating representing the actual reaction of the selected user to that item, the step of defining including entering, into an input device of the computing device, information representing the reaction of the selected user to items sampled by that user;

defining, for each item sampled by other users in the group, an actual scalar rating representing the actual reaction of each user to that item, including entering into the input device information representing the reaction of the other users to items sampled by each user;

successively pairing the selected user with other users in the group for whom have been defined actual scalar rating for at least some of the items sampled by the selected user to determine the difference in actual ratings for items sampled by both members of each successive pair;

designating a plurality of the other users as updating users and assigning a weighting value to each of the updating users based on the difference on actual ratings between that updating user and the selected user;

establishing an initial estimated rating for items not yet sampled by the selected user; and

applying the weighting values, for items not yet sampled by the selected user, to proportionally alter the difference between one of (a) the initial estimated rating for each item not yet sampled and (b) an estimated rating previously updated for each item not yet sampled by the selected user and the actual ratings of that item by the updating users to adjust the estimated ratings for the selected user to more closely predict the actual reaction of the user to that item.