Méthodes de stockage et d'interrogation de partitions musicales (Graph-based modelling and querying of music scores)

Sheet music scores have been the traditional way to preserve and disseminate Western classical music works for centuries. A sheet score is a complex semiotic object: in a single and compact layout, it combines a symbolic encoding of the music that must be produced with a sophisticated visual representation aiming at accurately representing the music content.

Nowadays, the content of sheet music scores can be encoded in digital formats. They provide a very detailed representation of music content expressed in the language of music notation, which is the most elaborate way of describing music at a symbolic level. Standard digital formats for encoding music scores rely on semi-structured data models (XML-based like MusicXML [5] and MEI [6,9]). Large collections of encoded music scores are now available, either as a result of long-running academic efforts [8], or as a result of the generalized production of music scores with editing software applications (e.g. MuseScore). Such collections are examples of digital libraries where the music content is described in a structured and well-organized way, ready to support sophisticated computer-based operations. Although the above-mentioned formats make possible to exchange music score data between software applications, they are poorly adapted to the ad hoc querying of the encoded score data, by users that lack programming skills.

Also, in the last few years, the database community has focused on graph databases to handle “complex” data [1,2,4,39]. Their basic purpose is to manage data natively modelled as graphs like social networks and biological, topological or bibliographic databases. As the content of an encoded score is composed of highly complex information, with connected items (a note follows another one, it belongs to a voice, it also belongs to a measure, etc.), its graph-based representation seems a relevant (intuitive) approach to leverage existing encodings to a true data model supporting ad hoc searches.

The goal of the PhD thesis is to study the modelling and the querying of music scores encoded in a graph-based representation. This includes the following tasks:

• -       studying and improving a graph-based data model previously proposed by the Shaman team [7], which is a first step towards a graph-based representation of music scores,

• -       studying the querying of such data through graph pattern queries, including the definition of a query language extension dedicated to the flexible querying of music scores (expressiveness, complexity, optimization mechanisms, benchmarks over a large volume of music score data), and

• -       in addition to their theoretical dimension above-mentioned, the contributions should be implemented in a digital library of scores that contains Traditional Breton folk music [10], for which user feedbacks should be collected.

The PhD student will be located in the Shaman team, in Lannion.

French version of the subject: https://people.irisa.fr/Virginie.Thion/OpenAccessResearch/IRISA-Shaman-…