LyricFind Corpus

Welcome to the LyricFind Corpus, developed at the Sound & Music Computing laboratory at the National University of Singapore with the very gracious support and partnership of LyricFind, a world leader in legal lyrics licensing and retrieval. In addition to providing the raw data which comprises the analyses in our ISMIR 2015 paper “Quantifying Lexical Novelty in Song Lyrics”, we are also pleased to provide 275,905 distinct lyrics in bag-of-words format (67.6 million total word instances), along with identifying lyric, artist, and album IDs that can be cross-referenced with the LyricFind ecosystem. We believe that this dataset marks the largest and cleanest set of lyrics in bag-of-words format yet available, although comparisons with the Million Song Dataset’s musiXmatch lyrics corpus are certainly warranted! The following six files are packaged a single large .zip file, available for download in either .txt. or .xlsx. For more details, please refer to "README.pdf". To reuse this dataset, please ensure the original research paper is cited appropriately. For details, please refer to Citation Field. Ellis, R.J., Xing, Z., Fang, J., & Wang, Y. (2015). Quantifying lexical novelty in song lyrics. Proceedings of the 15th International Conference on Music Information Retrieval. This dataset is also available at https://www.smcnus.org/lyrics/