Playlist generation via vector representation of songs

Abstract

This study proposes a song recommender system. The architecture is based on a distributed scalable big data framework. The recommender system analyzes songs a person listens to most and recommends a list of songs as a playlist. To realize the system, we use Word2vec algorithm by creating vector representations of songs. Word2vec algorithm is adapted to Apache Spark big data framework and run on distributed vector representation of songs to produce a playlist reflecting a person’s personal tastes. The performance results are evaluated in terms of hit rates at the end of the paper.