Systems and methods for improving accuracy in media asset recommendation models based on users' levels of enjoyment with respect to media assets

1. A method for training a model in order to improve the model's estimation of media asset ratings, the method comprising:

receiving, using communications circuitry, first preference information of a first plurality of users, wherein the first preference information is from a first data space and describes preferences of the first plurality of users with respect to a first plurality of media assets;

receiving, using the communications circuitry, second preference information, wherein:

the second preference information (1) is from a second data space that is different from the first data space, (2) describes preferences of a second plurality of users with respect to a second plurality of media assets, and (3) is computed using a different metric than a metric that the first preference information is computed using and wherein

the second data space is managed by a content provider that does not manage the first data space;

normalizing, using control circuitry, the first preference information and the second preference information such that both the first preference information and the second preference information are converted to a scheme on which a common metric may be applied;

determining, using the control circuitry, using the normalized first preference information and the normalized second preference information, a user's level of enjoyment with respect to a media asset based on the common metric, wherein the first preference information and the second preference information each comprise data describing the user's level of enjoyment of the media asset;

determining, using the control circuitry, an expected level of enjoyment that the user is expected to have with respect to the media asset as calculated by the model;

determining, using the control circuitry, an error value, wherein the error value is based on a comparison between the user's level of enjoyment and the expected level of enjoyment;

determining, using the control circuitry, trainable parameters of the model, wherein the trainable parameters of the model comprise updatable values used to improve accuracy of the expected level of enjoyment of the user with respect to the media asset as calculated by the model;

transmitting, using the communications circuitry, a request to a remote server hosting the model, for a plurality of weights stored at the remote server, where each weight of the plurality of weights corresponds to a trainable parameter of the model;

receiving, using the communications circuitry, the plurality of weights in response to the request;

updating, using the control circuitry, and based on the error value, the plurality of weights; and

transmitting to the remote server, using the communications circuitry, the updated plurality of weights and an instruction to replace the plurality of weights stored at the remote server with the updated plurality of weights.

2. The method of claim 1 , wherein normalizing the first preference information and the second preference information comprises generating a record for the media asset, wherein the record comprises preference information that is retrieved from both the first data space and the second data space, and wherein the retrieved preference information has been converted to the scheme on which the common metric may be applied.

3. The method of claim 1 , wherein determining, the user's level of enjoyment with respect to the media asset comprises:

identifying metadata associated with the media asset;

comparing the identified metadata associated with the media asset with metadata associated with a second media asset, wherein the second media asset is from the second plurality of media assets;

determining whether the metadata of the media asset matches the metadata of the second media asset; and

in response to determining that the metadata of the media asset matches the metadata of the second media asset, determining the user's level of enjoyment with respect to the media asset based on the first preference information associated with the first media asset and the second preference information associated with the second media asset.

4. The method of claim 1 , wherein determining, using the normalized first preference information and the normalized second preference information, the level of enjoyment that the user has with respect to the media asset further comprises:

calculating a first confidence value in the user's level of enjoyment of the media asset based on the first preference information, wherein the first confidence value indicates a first degree of accuracy of data from the first data space;

calculating a second confidence value in the user's level of enjoyment of the media asset based on the second preference information, wherein the second confidence value indicates a second degree of accuracy of data from the second data space;

determining a combined confidence value based on the first confidence value and the second confidence value; and

adjusting the level of enjoyment that the user has with respect to the media asset based on the combined confidence value.

5. The method of claim 4 , wherein the first confidence value is based on a number of users that consumed the media asset in the first data space.

6. The method of claim 4 , wherein determining the combined confidence value based on the first confidence value and the second confidence value further comprises:

determining a first degree of particularity, wherein the first degree of particularity is based on the first preference information, and wherein the first degree of particularity indicates a first degree of precision, in the first data space, of the user's level of enjoyment with respect to the media asset;

determining a second degree of particularity, wherein the second degree of particularity is based on the second preference information, and wherein the second degree of particularity indicates a second degree of precision, in the second data space, of the user's level of enjoyment with respect to the media asset;

calculating a combined particularity value based on the first degree of particularity and the second degree of particularity ; and

determining the combined confidence value based on the combined particularity value.

7. The method of claim 6 , wherein determining the combined confidence value based on the combined particularity value comprises:

calculating a weighted average of the first degree of particularity and the second degree of particularity.

8. The method of claim 1 , further comprising:

computing a derivative of a composition of both (1) a first function used to determine the level of enjoyment that the user has with respect to the media asset and (2) a second function used by the model to determine the expected level of enjoyment that the user is expected to have with respect to the media asset; and

updating the model based on the computed derivative.

9. The method of claim 8 , wherein updating the model based on the computed derivative comprises:

updating the plurality of weights based on the computed derivative.

10. A system for training a model in order to improve the model's estimation of media asset ratings, the system comprising:

communications circuitry configured to:

receive first preference information of a first plurality of users, wherein the first preference information is from a first data space and describes preferences of the first plurality of users with respect to a first plurality of media assets;

receive second preference information, wherein:

the second preference information (1) is from a second data space that is different from the first data space, (2) describes preferences of a second plurality of users with respect to a second plurality of media assets, and (3) is computed using a different metric than a metric that the first preference information is computed using and wherein

the second data space is managed by a content provider that does not manage the first data space;

transmit a request to a remote server hosting the model, for a plurality of weights stored at the remote server, where each weight of the plurality of weights corresponds to a trainable parameter of the model;

receive the plurality of weights in response to the request;

transmit to the remote server the updated plurality of weights and an instruction to replace the plurality of weights stored at the remote server with the updated plurality of weights; and

control circuitry configured to:

normalize the first preference information and the second preference information such that both the first preference information and the second preference information are converted to a scheme on which a common metric may be applied;

determine, using the normalized first preference information and the normalized second preference information, a user's level of enjoyment with respect to a media asset based on the common metric, wherein the first preference information and the second preference information each comprise data describing the user's level of enjoyment of the media asset;

determine an expected level of enjoyment that the user is expected to have with respect to the media asset as calculated by the model;

determine an error value, wherein the error value is based on a comparison between the user's level of enjoyment and the expected level of enjoyment;

determine trainable parameters of the model, wherein the trainable parameters of the model comprise updatable values used to improve accuracy of the expected level of enjoyment of the user with respect to the media asset as calculated by the model; and

update, based on the error value, the plurality of weights.

11. The system of claim 10 , wherein the control circuitry is further configured, when normalizing the first preference information and the second preference information, to generate a record for the media asset, wherein the record comprises preference information that is retrieved from both the first data space and the second data space, and wherein the retrieved preference information has been converted to the scheme on which the common metric may be applied.

12. The system of claim 10 , wherein the control circuitry is further configured when determining, the user's level of enjoyment with respect to the media asset, to:

identify metadata associated with the media asset;

compare the identified metadata associated with the media asset with metadata associated with a second media asset, wherein the second media asset is from the second plurality of media assets;

determine whether the metadata of the media asset matches the metadata of the second media asset; and

in response to determining that the metadata of the media asset matches the metadata of the second media asset, determine the user's level of enjoyment with respect to the media asset based on the first preference information associated with the first media asset and the second preference information associated with the second media asset.

13. The system of claim 10 , wherein the control circuitry is further configured, when determining, using the normalized first preference information and the normalized second preference information, the level of enjoyment that the user has with respect to the media asset, to:

calculate a first confidence value in the user's level of enjoyment of the media asset based on the first preference information, wherein the first confidence value indicates a first degree of accuracy of data from the first data space;

calculate a second confidence value in the user's level of enjoyment of the media asset based on the second preference information, wherein the second confidence value indicates a second degree of accuracy of data from the second data space;

determine a combined confidence value based on the first confidence value and the second confidence value; and

adjust the level of enjoyment that the user has with respect to the media asset based on the combined confidence value.

14. The system of claim 13 , wherein the first confidence value is based on a number of users that consumed the media asset in the first data space.

15. The system of claim 13 , wherein the control circuitry is further configured, when determining the combined confidence value based on the first confidence value and the second confidence value, to:

determine a first degree of particularity, wherein the first degree of particularity is based on the first preference information, and wherein the first degree of particularity indicates a first degree of precision, in the first data space, of the user's level of enjoyment with respect to the media asset;

determine a second degree of particularity, wherein the second degree of particularity is based on the second preference information, and wherein the second degree of particularity indicates a second degree of precision, in the second data space, of the user's level of enjoyment with respect to the media asset;

calculate a combined particularity value based on the first degree of particularity and the second degree of particularity ; and

determine the combined confidence value based on the combined particularity value.

16. The system of claim 15 , wherein the control circuitry further configured, when determining the combined confidence value based on the combined particularity value, to:

calculate a weighted average of the first degree of particularity and the second degree of particularity.

17. The system of claim 10 , wherein the control circuitry is further configured to:

compute a derivative of a composition of both (1) a first function used to determine the level of enjoyment that the user has with respect to the media asset and (2) a second function used by the model to determine the expected level of enjoyment that the user is expected to have with respect to the media asset; and

update the model based on the computed derivative.

18. The system of claim 17 , wherein the control circuitry is further configured, when updating the model based on the computed derivative, to:

update the plurality of weights based on the computed derivative.