Inferring attribute and item preferences
View Patent ↗User-preference datapoints are collected. At least some of these datapoints are associated with user-preference information about an item (e.g., a movie), and some of these datapoints are associated with user-preference information about an attribute (e.g., a movie genre or an actor). A profile of the user is created based, at least in part, on these datapoints. When a new datapoint is collected, the new datapoint is assigned a user-preference “score.” If, for example, the new datapoint does not come with an explicit user-preference rating, then the score is based on related item and attribute datapoints already in the profile. Depending upon the relationship of the new datapoint to the already existing datapoints, a confidence value is assigned to the user-preference score. The profile is then updated with the new datapoint along with its assigned score and confidence rating. The user profile can be used in performing any number of actions.
1. A method for a profile engine to perform an action based on an input datapoint, the method comprising:
providing, to the profile engine, user-preference scores for a plurality of previous datapoints, the plurality of previous datapoints comprising an item datapoint and an attribute datapoint, the input datapoint distinct from the plurality of previous datapoints;
assigning, by the profile engine, a user-preference score to the input datapoint based on a cross-domain inference,
wherein the cross-domain inference comprises an inference, based on a first genre preference for a genre associated with items in a first domain, of a second genre preference for a corresponding genre associated with items in a second domain different from the first domain, wherein the first domain and the second domain are selected from a group consisting of movies, programming, music, books, restaurants, and websites, and
wherein if the input datapoint is an item datapoint, then the assigning is based, at least in part, on a user-preference score for a previous attribute datapoint, else if the input datapoint is an attribute datapoint, then the assigning is based, least in part, on a user-preference score for a previous item datapoint;
based on a cross-domain inference, wherein the cross-domain inference comprises an inference, based on a first genre preference for a genre associated with items in a first domain, of a second genre preference for a corresponding genre associated with items in a second domain different from the first domain;
based, at least in part, on a comparison of the input datapoint with at least a subset of the plurality of previous datapoints, assigning, by the profile engine, a confidence value to the input datapoint, and lowering the confidence value assigned to the input datapoint based on the cross-domain inference;
based, at least in part, on the user-preference score and on the confidence value assigned to the input datapoint, performing, by the profile engine, an action.
2. The method of claim 1 wherein the input datapoint is selected from the group consisting of: a behavioral observation, a preference explicitly stated by a user, a preference explicitly stated by something other than the user, passive usage data, passive contextual data, and a statistical aggregation of behavioral data.
3. The method of claim 1 wherein the input datapoint is associated with an entity selected from the group consisting of: a person, a plurality of persons, a corporate entity, a social network of persons, a group of persons defined by a shared contextual attribute, and an entity defined by a collection rule.
4. The method of claim 3 wherein the plurality of the previous datapoints are associated with the entity.
5. The method of claim 1 wherein if the input datapoint is an item datapoint, then the assigning is further based on a user-preference score for a previous item datapoint, else if the input datapoint is an attribute datapoint, then the assigning is further based on a user-preference score for a previous attribute datapoint.
6. The method of claim 1 wherein performing is further based on a rule.
7. The method of claim 1 wherein performing is further based on a policy.
8. The method of claim 1 wherein the action comprises an element selected from the group consisting of: directing an advertisement, directing a notification, initiating an interaction, providing a recommendation, choosing a message to send, providing a prediction, and mapping a behavioral pattern.
9. The method of claim 1 further comprising:
presenting, to an entity associated with the input datapoint, the user-preference score assigned to the input datapoint;
receiving, from the entity, an amendment to the user-preference score for the input datapoint; and
amending the confidence value for the input datapoint.
10. The method of claim 9 further comprising:
based, at least in part, on the amended user-preference score for the input datapoint and on the amended confidence value for the input datapoint, performing, by the profile engine, an action.
11. The method of claim 1 wherein assigning, by the profile engine, the user-preference score to the input datapoint further comprises:
applying a rule for ignoring the input datapoint based on privacy reasons.
12. A profile engine configured for performing an action based on an input datapoint, the profile engine comprising:
a memory configured for storing user-preference scores for a plurality of previous datapoints, the plurality of previous datapoints comprising an item datapoint and an attribute datapoint, the input datapoint distinct from the plurality of previous datapoints; and
a processor operatively connected to the memory and configured for:
assigning a user-preference score to the input datapoint based on a cross-domain inference,
wherein the cross-domain inference comprises an inference, based on a first genre preference for a genre associated with items in a first domain, of a second genre preference for a corresponding genre associated with items in a second domain different from the first domain, wherein the first domain and the second domain are selected from a group consisting of movies, programming, music, books, restaurants, and websites, and
wherein if the input datapoint is an item datapoint, then the assigning is based, at least in part, on a user-preference score for a previous attribute datapoint, else if the input datapoint is an attribute datapoint, then the assigning is based, least in part, on a user-preference score for a previous item datapoint;
based, at least in part, on a comparison of the input datapoint with at least a subset of the plurality of previous datapoints, assigning a confidence value to the input datapoint, and lowering the confidence value assigned to the input datapoint based on the cross-domain inference; and
based, at least in part, on the user-preference score and on the confidence value assigned to the input datapoint, performing an action.
13. The profile engine of claim 12 wherein the profile engine is selected from the group consisting of: a personal electronics device, a mobile telephone, a personal digital assistant, a tablet computer, a gaming console, a compute server, and a coordinated group of compute servers.
14. The profile engine of claim 12 wherein assigning, by the profile engine, the user-preference score to the input datapoint further comprises:
applying a rule for ignoring the input datapoint based on privacy reasons.