Results

See below left for the sample of playlist generated for romantic mood, and right for the sample of playlist generated for melancholic mood. Analysis of the recommended songs can be see further at the bottom.

Result Interpretation

In this algorithm, we would recommend songs based on mood. We divided songs into categories, based on their lyrics. And when user selects a specific mood, the model would be able to recommend songs in the selected mood category. (The interactive user interface is to be done for the future work, please check out the "Conclusions / Future Work" page for more information. )

Here, we have two examples of the mood categories: Romantic and Melancholy.

For instance, in the Romantic category, we have "Can't Help Falling in Love", "(I know) I'm Loosing You", "What's Love Gotta Do With It". And in the Melancholy category, we have "Me and My Broken Heart", "Never Again", "Crying in the Rain". From the titles, we can see that most songs agree with the mood categories.

Melancholy section is also interesting to see, especially seeing songs like "Sing For The Moment" by Eminem and "Never Again" by Justin Timberlake which fits the melancholic feeling perfectly. Most of the other songs acoustically represent the melancholic aspect too. "Seventeen Forever" by Metro Station sounds more upbeat just by listening to it, but looking into the lyrics, it actually does have melancholic meaning and connotation attached to the song itself, so it is not necessarily a bad recommendation.

Similar to the previous recommended songs, we have again asked several users (n = 5) to play the original playlist of songs and the recommended songs as a rudimentary user testing, and they all agreed that acoustically, they both share similar mood of the song which further validates the effectiveness of the recommended songs from our model.