Rating and novelty decay
Application servers and methods of operating the same are provided for generating a personalized recommendation of items to a first user. An updated rating value N(t) of an initial rating value N(0) is determined for each rated item based on an age of each rating. The updated rating value N(t) is based on a difference between the value of the initial rating N(0) and a neutral rating value R, and on a predetermined half-life of the ratings. The updated rating value N(t) converges towards the neutral rating value R with an increase in the age t for each rating.
1. An application server in a recommender system, comprising:
a processor; and
a memory coupled to the processor that comprises a computer program that when executed by the processor causes the processor to perform operations comprising:
calculating an initial rating value N(0) for each of a plurality of items rated by a user;
determining a neutral rating value R for each of the plurality of items rated by the user; and
calculating a value of an updated rating N(t) for each of the rated items, based on a difference between the initial rating value N(0) and the neutral rating value R and on a predetermined half-life (t½[rating]) of each one of the ratings, the updated rating value N(t) converging towards the neutral rating value R with an increase in the age t of each one of the ratings;
wherein a converging value of the updated rating N(t) depends on a value of e −(ln2/t1/2[rating])×t .
2. An application server according to claim 1 , wherein calculating the initial rating value N(0) comprises calculating the initial rating value N(0) as a function of at least one input parameter.
3. An application server according to claim 1 , wherein the neutral rating value R corresponds to an average rating for the rated items associated with each user.
4. An application server according to claim 1 , wherein the neutral rating value R corresponds to an average of different user ratings associated with each of the rated items.
5. An application server according to claim 1 , wherein a value of a half-life for each of the items t½[item] depends on an item type.
6. An application server, comprising:
a recommender system provided with a rating updating unit configured to calculate an initial rating value N(0) for each of a plurality of items rated by a user and to calculate an update of an initial rating value (N(0)) for each of a plurality of user rated items based on an age (t) of each one of the ratings, the rating updating unit comprising:
a neutral rating unit configured to determine a neutral rating value (R) for each of the plurality of user rated items; and
an updated rating value unit configured to calculate an updated rating (N(t)) for each of the rated items, based on the difference between the initial rating value N(0) and the neutral rating value R and on a predetermined half-life (t½[rating]) of each one of the ratings, the updated rating value N(t) converging towards the neutral rating value with an increase in the age t of each one of the ratings;
wherein a converging value of the updated rating N(t) depends on a value of e −(ln2/t1/2[rating])×t .
7. An application server according to claim 6 , wherein the rating updating unit is further configured to calculate the initial rating value N(0) as a function of at least one input parameter.
8. An application server according to claim 6 , wherein the neutral rating value R corresponds to an average rating for the rated items associated with each user.
9. An application server according to claim 6 , wherein the neutral rating value R corresponds to an average of different user ratings associated with each of the rated items.
10. An application server according to claim 6 , further comprising a database for storing parameters used by the rating updating unit.