Data Collection

For collecting Irish Folk Tunes, I used the following resources -

• Henrik Norbeck's ABC Book -> 365 Double Jigs from Capt. Francis ONeil's Dance Music of Ireland
• TheSession Data Dump -> Collection of Irish tunes from thesession.org

From these resources, I was able to obtain a decent number of tunes. However, ABC tunes are usually packaged with a lot of metadata which is irrelevant within the context of deep learning. So after downloading the raw data, I extract key, rhythm, meter and the ABC tune itself, and store it within a CSV file. Further, I normalize the key names and remove whitespaces and newlines across all the tunes.

I create different CSV files to contain different sub-collections of tunes. The sub-collections of tunes that I use are provided below.

• 365 Double Jigs - ONeil's Dance Music of Ireland
• All tunes in 6/8 Meter from TheSession Data Dump
• All tunes from TheSession Data Dump

These CSV files are further used for creation of datasets by feeding them into the preprocessing pipeline. You can read more about the next steps here.