Lyrical Influence Networks Dataset (LIND)

This dataset contains three directed networks (graphs) that encode similarity between song genres, artists, and writers based on phrases shared across song lyrics over time. Each graph is presented in a separate .graphml file. The graphs contain nodes for 214 genres, 42,802 artists, and 95,349 writers respectively, with shared phrase relationships constructed from the intact lyrics of 554,206 songs. Information regarding the construction of the networks, as well as example analyses, are provided in the accompanying published ISMIR paper (Atherton & Kaneshiro, 2016). The intact lyrics used for construction of the networks were obtained from LyricFind (http://www.lyricfind.com/) via signed research agreement; the networks are published here with permission from LyricFind. Graph files are available for download in compressed (.zip) format and range in size from 342 KB - 776.5 MB compressed to 2.7 MB - 9.04 GB uncompressed.