From melodic note sequences to pitches using word2vec
This is an incremental application of an existing NLP method to music data, potentially aiding in music analysis or generation tasks.
The study tackled the problem of capturing pitch information from melodic note sequences by applying word2vec, treating notes as words, and achieved a multiple correlation coefficient of approximately 0.80 between semantic vectors and pitches.
Applying the word2vec technique, commonly used in language modeling, to melodies, where notes are treated as words in sentences, enables the capture of pitch information. This study examines two datasets: 20 children's songs and an excerpt from a Bach sonata. The semantic space for defining the embeddings is of very small dimension, specifically 2. Notes are predicted based on the 2, 3 or 4 preceding notes that establish the context. A multivariate analysis of the results shows that the semantic vectors representing the notes have a multiple correlation coefficient of approximately 0.80 with their pitches.