Interinstitutional work is one of the final projects presented in PCS
The presentation of the final projects of students graduated in 2015 was December 16th, at the Electrical Engineering building. The students, who had already shown their projects working the day before, now had to present their projects to professors and other spectators. More than discussing the background theory, they also had to explain how the theory was applied and explored.
Among the works presented, there was the work of Alexandre Braga Saldanha, a Computer Science student at the Universidade Federal de São Carlos, and Nathália Marques de Oliveira, a Computer Engineer student from PCS. It was a singular and interesting work due to its information exchange and proximity of students from different universities, courses, and cities.
The students worked with the dynamic recommendation of music based on grouping. With the popularity of audio digital format (46% of the music market is digital media), many people have created large personal collections of music. These collections, which may have hundreds or thousands of music, reflect the musical interest of its listeners. However, each person may daily vary its humor, emotional state, and context, what impact at her musical choice. If the collection is too large, manually selecting the playlist may be a hard work. If the person uses the shuffle function, the playlist generated has no relationship between the tracks, as the shuffle chooses tracks randomly.
Considering this problem, Alexandre and Nathália developed a dynamic music recommendation system, which allows users listening to musics from her collection that are chosen based on her interested shown at that moment.
The recommendation is based on grouping similar music, generated from characteristics obtained from the music tracks that the user has. While the user is listening to her music, the system uses these groupings and the implicit feedback from the user (interpreting tracks listened as approved and tracks skipped as rejected) to recommend the next music to be listened. For example, if the user skipped reggae music, but listened to pop music, the system would recommend more pop music and less reggae music. Therefore, the choice is not random as in a shuffle.
Even though the test of the system had some problems, such as analyzing a small number of users (only two) and limiting the user actions (approve or reject) only after the track ended (what made the process slow), Nathália and Alexandre fulfilled their goal, applying concepts learned during their course. More than executing a very interesting interinstitutional project, they also created a system that, if improved, can be very useful in today’s digital media reality.
Gabriel Campos e Nelson Niero | Jornalismo Júnior
Image | Nilton Araújo do Carmo