Recommender system with AD-HOC, dynamic model composition
View Patent ↗One embodiment of the present invention provides recommender system for generating a recommendation of an item by combining a set of utility models adaptively to facilitate a decision-making process. The system includes a utility model database containing the set of utility models and a query module for receiving at least one query about the item from a querying entity. The system also includes a rule engine to specify a subset of utility models to be applied to the item and to specify a weight function of the specified utility models. Further included in the system is a set generator coupled to the utility model database, the query module, and the rule engine. The set generator computes a set of ratings by applying each of the utility model in the subset to the item and generates an overall rating for the item based on the weight function. The system further a communication module to return the overall rating.
1. A recommender system for generating a recommendation of an item by combining a set of utility models adaptively to facilitate a decision-making process, the system comprising:
a utility model database containing the set of utility models;
a query module for receiving at a query about the item from a querying entity;
a rule engine to:
specify two or more utility models to be applied to the item,
determine a weight for each utility model, and
specify a weight function of the utility models based in part on the determined weights;
a set generator to:
compute a set of ratings by applying each utility model to the item, and
generate an overall rating for the item based in part on the weight function; and
a communication module to return the overall rating.
2. The recommender system of claim 1 , wherein the rule engine includes a user context module to receive contextual and/or preferential information associated with the querying entity.
3. The recommender system of claim 2 , wherein the contextual information includes at least one of:
a current time of day;
a current weather condition;
a location associated with the querying entity;
a movement associated with the querying entity;
a calendar entry reflecting a user's past, current, and/or future schedule;
a set of messaging data comprised of text extracted from, among others, emails and/or instant messages, where one or more users discuss interests and/or plans; and
a reading history which involves content extracted from web pages or other sources a user has accessed, which content reflects the user's interests and/or tastes.
4. The recommender system of claim 1 , wherein the rule engine selects the subset of utility models based on the contextual and/or preferential information received by the user context module.
5. The recommender system of claim 1 , wherein the rule engine associates a decision made by the querying entity with the weight function to adjust the weight function adaptively.
6. The recommender system of claim 1 , wherein the weight function of the specified utility models is constructed by employing a heuristic approach.
7. The recommender system of claim 1 , wherein the weight function of the specified utility models is constructed by employing a statistical approach.
8. A method for generating a recommendation of an item by combining a set of utility models adaptively to facilitate a decision-making process, the method comprising:
maintaining the set of utility models;
receiving a query about the item from a querying entity;
specifying two or more utility models to be applied to the item;
determining a weight for each utility model;
specifying a weight function of the utility models based in part on the determined weights;
computing a set of ratings by applying each utility model to the item;
generating an overall rating for the item based in part on the weight function; and
returning the overall rating.
9. The method of claim 8 , further comprising receiving contextual and/or preferential information associated with the querying entity.
10. The method of claim 9 , wherein the contextual information includes at least one of:
a current time of day;
a current weather condition;
a location associated with the querying entity;
a movement associated with the querying entity;
a calendar entry reflecting a user's past, current, and/or future schedule;
a set of messaging data comprised of text extracted from, among others, emails and/or instant messages, where one or more users discuss interests and/or plans; and
a reading history which involves content extracted from web pages or other sources a user has accessed, which content reflects the user's interests and/or tastes.
11. The method of claim 8 ,
wherein specifying the subset of utility models comprises selecting the utility models based on the contextual and/or preferential information received by the user context module. .
12. The method of claim 8 , further comprising associating a decision made by the querying entity with the weight function to adjust the weight function adaptively.
13. The method of claim 8 , further comprising employing a heuristic approach to construct the weight function.
14. The method of claim 8 , further comprising employing a statistical approach to construct the weight function.
15. A computer readable storage medium storing instructions which when executed by a computer cause the computer to perform a method for generating a recommendation of an item by combining a set of utility models adaptively to facilitate a decision-making process, the method comprising:
maintaining the set of utility models;
receiving at least one query about the item from a querying entity;
specifying two or more utility models to be applied to the item;
determining a weight for each utility model;
specifying a weight function of the specified utility models based in part on the determined weights;
computing a set of ratings by applying each of the utility model to the item;
generating an overall rating for the item based in part on the weight function; and
returning the overall rating.
16. The computer readable storage medium of claim 15 , wherein the method further comprises receiving contextual and/or preferential information associated with the querying entity.
17. The computer readable storage medium of claim 16 , wherein the contextual information includes at least one of:
a current time of day;
a current weather condition;
a location associated with the querying entity;
a movement associated with the querying entity;
a calendar entry reflecting a user's past, current, and/or future schedule;
a set of messaging data comprised of text extracted from, among others, emails and/or instant messages, where one or more users discuss interests and/or plans; and
a reading history which involves content extracted from web pages or other sources a user has accessed, which content reflects the user's interests and/or tastes.
18. The computer readable storage medium of claim 15 ,
wherein specifying the subset of utility models comprises selecting the utility models based on the contextual and/or preferential information received by the user context module.
19. The computer readable storage medium of claim 15 , wherein the method further comprises associating a decision made by the querying entity with the weight function to adjust the weight function adaptively.
20. The computer readable storage medium of claim 15 , wherein the method further comprises employing a heuristic approach to construct the weight function.
21. The computer readable storage medium of claim 15 , wherein the method further comprises employing a statistical approach to construct the weight function.