A man-made intelligence (AI) model has emerged that may predict Grammy winners. The model has actually been known to appropriately predict all nine major Grammy category winners over the past three years.
Futurity reported on the fifth (local time) that researchers at Latest York University developed a model trained on data from Grammy Award winning songs, and thru this, they were capable of appropriately predict all the Grammy winning songs from the past three years.
The research team built a dataset with about 250 nominated songs in three categories, including Song of the 12 months, Record of the 12 months, and Rap Song of the 12 months, from 2004 to 2020, and trained the model.
The model was then asked to play a complete of 27 songs, three from each category from 2021 to 2023 that weren’t included within the training, and find the winner.
Because of this, the model revealed that it accurately identified all nine winners.
The authors explained that a number of the model’s predictions were different from those made by the betting sites. For instance, Bonnie Raitt’s “Just Like That,” the 2023 Song of the 12 months, and H.E.R.’s “I Can’t Breathe,” the 2021 winner, were each given low odds by the betting sites on the time.
It was revealed that big data equivalent to Billboard rankings and Google search volumes were used to coach the model.
Additionally they explained that they precisely analyzed the musical characteristics using Spotify data, namely ▲whether it’s acoustic or electronic ▲whether it’s suitable for dancing ▲the energy level of the song ▲the ratio of vocals and instrumental performances ▲lyrics, etc.
Specifically, for the lyrics, they said they calculated specific words and the emotional tone conveyed by them, and even swear words, through a natural language processing algorithm.
“Identifying award-winning artistic endeavors is clearly a subjective process, made much more complicated by the key ballot process,” said Anna Barry, an assistant professor at Latest York University’s Institute for Mathematical Sciences.
Nonetheless, he said, “If we consider every little thing concerning the song itself, from its composition to its popularity, we will accurately predict the winner.”
He added that the tool could also help discover latest artists and trends that may develop into popular in the long run. “I hope this tool will make clear music that otherwise may not have been discovered,” he said.
This study was presented on the International Conference on Big Data Analytics (ICBDA) in March and was published by the Institute of Electrical and Electronics Engineers (IEEE). Recently published in Explorer JournalOkay.
Reporter Im Dae-jun ydj@aitimes.com
