Methods and apparatus for predicting and selectively collecting preferences based on personality diagnosis

• US 6,655,963 B1
• Filed: 07/31/2000
• Issued: 12/02/2003
• Est. Priority Date: 07/31/2000
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Pennock, David M., Horvitz, Eric J.
• Primary Examiner(s): Harrison, Jessica
• Assistant Examiner(s): ADAMS, CHANDA L
• Application Number:
  US09/629,217
• Time in Patent Office: 1,219 Days
• Field of Search: 434/236, 707/10, 707/7, 705/26, 705/10, 705/1
• US Class Current: 434/236
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0204 Market segmentation
• CPC Class Codes: G06Q 30/0625 Directed, with specific intent or strategy
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: H04N 21/4668 for recommending content, e.g. movies
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

A new recommendation technique, referred to as “personality diagnosis”, can be seen as a hybrid between memory-based and model-based collaborative filtering techniques. Using personality diagnosis, all data can be maintained throughout the processes, new data can be added incrementally, and predictions have meaningful probabilistic semantics. Each entity'"'"'s reported attributes can be interpreted as a manifestation of their underlying personality type. Personality type can be encoded simply as a vector of the entity'"'"'s “true” values for attributes in the database. Given an active entity'"'"'s known attribute values, the probability that they have the same personality type as every other entity can be determined. Then, the probability that they will have a given value for a valueless attribute can then be determined based on the entity'"'"'s personality type. The probabilistic determinations can be used to determine expected value of information.

Claims

1. A method for using a machine to predict a value of an attribute, having no assigned value, of an active entity, the method comprising:

a) accepting, with the machine, values of attributes of a number of other entities;

b) generating, with the machine, for each of the other entities, a probability that the active entity'"'"'s true personality type is that of the current other entity;

c) determining, with the machine, for each possible value of the attribute having no assigned value, a probability that the active entity values the attribute with the current possible value based, at least in part, on the probabilities that the active entity has a true personality type which is the same as that of the other entities as generated in act (b); and

d) selecting, with the machine, from among the possible values of the attribute having no assigned value, the possible value with the maximum probability determined in act (c) to generate a predicted value.

2. The method of claim 1 wherein the active entity is a person, each of the other entities is a person, each of the attributes is an item, and each of the values is an item rating.

3. The method of claim 1 wherein the active entity is a person, each of the other entities is a person, each of the attributes is an item selected from a group of items consisting of (a) textual content, (b) video content, (c) audio content, (d) image content, (e) multi-media content, (f) a service, (g) a consumer good, (h) a business good, (i) clothing, and (j) a financial instrument, and each of the values is an item rating.

4. The method of claim 1 wherein the act of generating, for each of the other entities, a probability that the active entity'"'"'s true personality type is that of the current other entity includes determining, with the machine, for each of the attributes, a probability that the active entity will truly value the attribute with the same value as the current other entity given the active entity'"'"'s values of other attributes.

5. The method of claim 1 wherein the act of generating, for each of the other entities, a probability that the active entity'"'"'s true personality type is that of the current other entity is based on

i) for each attribute, a probability that the active entity values the attribute given that the active entity'"'"'s true value of the attribute is the same as that of the current other entity, and

ii) a probability that the active entity'"'"'s true personality type is that of the current other entity.

6. The method of claim 5 wherein the act of determining, for each attribute, a probability that the active entity values the attribute given that the active entity'"'"'s true value of the attribute is the same as that of the current other entity is done in accordance with the expression:

 

e<sup>−

(x−

y)2/2σ

2</sup>,

where x is the active entity'"'"'s value of the attribute, y is the current other entity'"'"'s value of the attribute, and σ

is a user definable parameter.

7. The method of claim 5 wherein, in the act of determining a probability that the active entity'"'"'s true personality type is that of the current other entity includes setting, with the machine, the probability that the active entity'"'"'s true personality type is that of the current other entity is set to 1/n, where n is the number of other entities.

8. The method of claim 1 wherein the act of determining, for each possible value of the attribute having no assigned value, a probability that the active entity values the attribute with the current possible value is based on

i) for each personality type, a probability that the active entity values the current unknown attribute with the current value given that the active entity is of the current personality type, and

ii) a probability that the active user is of the current personality type.

9. The method of claim 8 wherein, in the act of determining a probability that the active entity'"'"'s true personality type is that of the current other entity includes setting, with the machine, the probability that the active entity'"'"'s true personality type is that of the current other entity is set to 1/n, where n is the number of other entities.

10. The method of claim 1 further comprising:

e) soliciting values of attributes from the active entity.

11. The method of claim 10 further comprising:

f) estimating an expected value, to the attribute value prediction, of having a value of an attribute, wherein the act of soliciting values of attributes from the active entity is controlled by function of the expected value estimated and a cost function of soliciting the value of the attribute from the active entity.

12. The method of claim 1 further comprising:

e) estimating an expected value, to the attribute value prediction, of having a value of an attribute, wherein the act of generating, with the machine, for each of the other entities, a probability that the active entity'"'"'s true personality type is that of the current other entity considers an incomplete set of all attributes, wherein the members of the incomplete set of all attributes are selected based on a function of the expected value estimated and a cost function of including the value of the attribute in the incomplete set.

13. A machine readable medium having instructions which, when executed by a machine, effect the method of claim 1.

14. A method for using a machine to generate, for each of a number of personality types defined by a plurality of other entities, probabilities that an active entity is that personality type, the method comprising:

a) accepting, with the machine, values of attributes associated with a number of other entities; and

b) for each of the other entities,

i) determining, with the machine, for each attribute, a probability that the active entity values the attribute given that the active entity'"'"'s true value of the attribute is the same as that of the current other entity, and

ii) determining, with the machine, a probability that the active entity'"'"'s true personality type is that of the current other entity.

15. The method of claim 14 wherein the act of determining, for each attribute, a probability that the active entity values the attribute given that the active entity'"'"'s true value of the attribute is the same as that of the current other entity is done in accordance with the expression:

16. The method of claim 14 wherein the act of determining a probability that the active entity'"'"'s true personality type is that of the current other entity includes setting, with the machine, the probability that the active entity'"'"'s true personality type is that of the current other entity to 1/n, where n is the number of other entities.

17. A machine readable medium having instructions which when executed by a machine, effect the method of claim 14.

18. A method for using a machine to determine, for each possible value of an attribute having no assigned value, a probability that an active entity values the attribute with the current possible value, the method comprising:

a) accepting, with the machine, for each of a number of personality types, a probability that the active user is of the current personality type;

b) determining, with the machine, a probability that the active entity values the attribute with the current possible value based on

i) for each personality type, a probability that the active entity values the current unknown attribute with the current value given that the active user is of the current personality type, and

ii) a probability that the active user is of the current personality type.

19. The method of claim 18 wherein the probability that the active entity'"'"'s true personality type is that of the current other entity is set to the probability that the active entity'"'"'s true personality type is that of the current other entity to 1/n, where n is the number of other entities.

20. A machine readable medium having instructions which, when executed by a machine, effect the method of claim 18.

21. An apparatus for predicting the value of an attribute of an active entity, the apparatus comprising:

a) a personality type generator for generating, for each of a plurality of personality types, a probability that the active entity is of the current personality type; and

b) an attribute value predictor for predicting the value of the attribute of the active entity based on the each of the probabilities that the active entity is of each of the personality types.

22. The apparatus of claim 21 wherein the personality type generator has access to values of attributes of each of a plurality of other entities.

23. The apparatus of claim 21 further comprising:

c) means for soliciting values of attributes from the active entity.

24. The apparatus of claim 23 further comprising:

d) means for estimating an expected value to the attribute value predictor of having a value of an attribute, wherein the means for soliciting values of attributes from the active entity are controlled by function of the expected value estimated and a cost function of soliciting the value of the attribute from the active entity.