Information processing device for musical score data

An information processing method realized by a computer, the information processing method comprising:

• generating variability data, which are change variables that represent a factor that alters a performance of a musical piece, and which follow a specific probability distribution, by inputting first training data, which include performance data that represent the performance of the musical piece, into a first provisional model;
• generating estimation data that represent the performance of the musical piece that reflects a change caused by the factor, by inputting second training data, which include musical score data that represent a musical score of the musical piece and the variability data generated by the first provisional model, into a second provisional model;
• and
• updating a plurality of coefficients that define the first provisional model and a plurality of coefficients that define the second provisional model, such that the performance data of the first training data and the estimation data generated by the second provisional model approach to each other, and such that a probability distribution of the variability data generated by the first provisional model approaches a specific target distribution.

An information processing device comprising:

• an electronic controller configured to execute
• generating variability data, which are change variables that represent a factor that alters a performance of a musical piece, and which follow a specific probability distribution, by inputting first training data, which include performance data that represent the performance of the musical piece, into a first provisional model,
• generating estimation data that represent the performance of the musical piece that reflects a change caused by the factor, by inputting second training data, which include musical score data that represent a musical score of the musical piece and the variability data generated by the first provisional model, into a second provisional model, and
• updating a plurality of coefficients that define the first provisional model and a plurality of coefficients that define the second provisional model, such that the performance data of the first training data and the estimation data generated by the second provisional model approach to each other, and such that a probability distribution of the variability data generated by the first provisional model approaches a specific target distribution.