Deezer, the researchers taught a neural network to determine the mood of the songs

Исследователи Deezer научили нейронную сеть определять настроение песен

Researchers service, stream music, Deezer has developed artificial intelligence that can determine the mood and intensity of the music. The corresponding work was published in arXiv.org.

To determine the mood of the tracks the researchers took into account how the audio and lyrics. First, they emit audible signals in the neural network together with the models that rekonstruiruet linguistic context of words.

Then, in order to teach the AI gets the Million Song database of metadata over one million contemporary songs. In particular, experts from Dezeer involved database Last.fm, which tracks are assigned identifiers of the more than 500 thousand unique tags.

Million Song database contains only metadata, and therefore the researchers linked information on a directory by Dezeer identifiers as song title, album and artist. About 60 percent of the data set (644 18 track) were used for training the AI, and the remaining 40 percent will be used for verification and further testing.

This work will help to explore how music, lyrics and mood of the songs correlate with each other.

Source

Оставьте комментарий

Ваш адрес email не будет опубликован. Обязательные поля помечены *