BMdataset: A Musicologically Curated LilyPond Dataset
For symbolic music researchers, this work demonstrates that small, high-quality curated datasets can be more effective than large noisy corpora for music understanding tasks, challenging the prevailing assumption that more data is always better.
The authors introduce BMdataset, a small but expertly curated dataset of 393 LilyPond scores, and show that fine-tuning a CodeBERT-based model (LilyBERT) on this dataset outperforms continuous pre-training on a much larger but noisier corpus (PDMX) for composer and style classification, achieving 84.3% composer accuracy when combined with broad pre-training.
Symbolic music research has relied almost exclusively on MIDI-based datasets; text-based engraving formats such as LilyPond remain unexplored for music understanding. We present BMdataset, a musicologically curated dataset of 393 LilyPond scores (2,646 movements) transcribed by experts directly from original Baroque manuscripts, with metadata covering composer, musical form, instrumentation, and sectional attributes. Building on this resource, we introduce LilyBERT (weights can be found at https://huggingface.co/csc-unipd/lilybert), a CodeBERT-based encoder adapted to symbolic music through vocabulary extension with 115 LilyPond-specific tokens and masked language model pre-training. Linear probing on the out-of-domain Mutopia corpus shows that, despite its modest size (~90M tokens), fine-tuning on BMdataset alone outperforms continuous pre-training on the full PDMX corpus (~15B tokens) for both composer and style classification, demonstrating that small, expertly curated datasets can be more effective than large, noisy corpora for music understanding. Combining broad pre-training with domain-specific fine-tuning yields the best results overall (84.3% composer accuracy), confirming that the two data regimes are complementary. We release the dataset, tokenizer, and model to establish a baseline for representation learning on LilyPond.