DATeCH 2017: Crowdsourced Transcription of Handwritten Music Scores


Author: Manuel Burghardt

I am currently attending the DATeCH 2017 conference in Göttingen, which is fully packed with interesting talks on „Digital Access to Textual Cultural Heritage“. The venue is right in the heart of the „Niedersächsische Staats- und Universitätsbibliothek Göttingen“, which is quite a decent place for a conference 🙂

Talk: Crowdsourced Transcription of Handwritten Music Scores

I will be presenting a research project that was created together with Sebastian Spanner.

Allegro: User-centered Design of a Tool for the Crowdsourced Transcription of Handwritten Music Scores

In this paper, we describe the challenge of transcribing a large corpus of handwritten music scores. We conducted an evaluation study of three existing optical music recognition (OMR) tools. The evaluation results indicate that OMR approaches do not work well for our corpus of highly heterogeneous, handwritten music scores. For this reason, we designed Allegro, a web-based crowdsourcing tool that can be used to transcribe scores. We relied on a user-centered design process throughout the development cycle of the application, to ensure a high level of usability. The interface was designed in a way it can be used intuitively, even by novices of musical notation. We describe the core features of Allegro and the basic transcription workflow. A first public beta test with 18 users shows that a crowdsourced transcription approach via the Allegro tool is a viable option for the encoding of our corpus of folk songs.

Related research from our group

Interesting tools mentioned in other DATeCH talks

  • BlackLab (corpus retrieval engine based on Apache Lucene):
  • Cosycat (Collaborative Synchronized Corpus Annotation Tool):
  • Vowpal Wabbit (fast learning algorithm):

*Gotta love those names 🙂

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert