Crowdsourced Digital Humanities Initiatives and Liturgical Chant

While optical music recognition (OMR) offers new avenues for the indexing and encoding of liturgical chant melodies, the process currently requires significant human effort and time. Recent advances (such as the Rodan workflow) have involved the use of machine learning to expedite and automate the process of chant identification, classification, and indexing. Machine learning can only work with an OMR system when effectively coordinated with human labour, however. That coordination depends on planning out and adapting individual stages of the workflow, and on the complexity of the source. Working with two Low Countries antiphoners from the Premonstratensian order (B-Gu, HS BKT.006 and B-AVna, Ms. IV.413), our trial tested various ways to improve and simplify the workflow. In particular, this trial tested crowdsourcing as a suitable compromise to expedite the text indexing and correction processes, stages not currently ready for machine learning integration. While complete machine learning integration remains a long-term goal for any OMR service–and particularly so for complex and non-standardized chant sources–our trial demonstrated the practicality of crowdsourced initiatives for specific stages of the workflow.