An Intelligent Music Playlist Generator Based on Time Parameter with Artificial Neural

An intelligent music playlist generator based on time parameter with artificial neural network
Ning-Han Liu , Shu-Ju Hsieh, Cheng-Fa Tsai.

This paper proposes an intelligent system that provides a suitable playlist to the user depending on the time the user listens to the music.In order to represent the musical piece this system extracts features from two types music files using artificial neural network (ANN). Thisplaylist generator is claimed to adjust to the user's daily activities in order to generate the appropriate music to suit the user’s current activity.

This system represents the characteristics of music from features extracted out of both the music’s symbolic form and wave data. The kernel of this system is based on a modified ANN. The user’s music rating history and the associated time stamps in the user’s profile constitute the training data of the modified artificial neural networks.

The user's preferences are recorded with a time stamp, unlike previous playlist generation systems that only record the playing preference. The ANN is trained by features of music that the user has listened to and the associated time parameter through the personal playlist generator training module. After the personal ANN is built, the user is then provided with a personalized playlist depending on the time parameter and the user’s preference of music.This structure includes two ANNs: one termed as the short term ANN which indicates the user’s recent behavior[1] to the music selection, while the other ANN is termed as the long term ANN which is used to re- cord the user’s long term preference in music.

The paper also addressed a problem named Cold Start (insufficient training data results in a playlist that is generated no different to a random selection)which was prominent in previous similar works. This system tries to minimize it using a collaborative method in which users are asked to fill a questionnaire during registration which helps identify possible similarity among different users. The system tries to suggest playlist using the data from similar users.

Knowledge of the Environment:

Programmed knowledge:
Features of different genre of music
Sensed knowledge:
User’s preference change with time
Affect of action:
User gets automatically generated playlist

Goal:
Generate automatic playlist based on users listening behavior over time

Knowledge about goal:
features of audio files/music

Utility function:
Short term ANN (recent behavior) and Long term ANN (long term preferences),
Pitch interval entropy extraction

Details of environment:

Accessibility:
Not fully accessible. Only supports 2 types of file and pop music.
Determinism:
Non-deterministic
Non-episodic:
Current music preference affects future action

Dynamic, discrete
Reference Chen, H.-C., & Chen, A. L. P. (2005).A music recommendation system based on music and user grouping. Intelligent Information Systems, 24(2/3), 113–132.