Mini-projects networking meeting
On Wednesday we held our networking meeting for those who have expressed an interest in applying for one of the Transforming Musicology mini-projects. It was hosted at the Lancaster Institute for Contemporary Arts (part of the University of Lancaster) by Alan Marsden, a co-investigator on the project and responsible for the mini-projects component. As well as Alan, Tim Crawford, Christophe Rhodes, Kevin Page, Ben Fields, and I were all present representing the project team.
The mini-projects were conceived as a way of engaging the musicological community at large with the project. This is vitally important: to make a credible case for a digital transformation of musicology we need to have the interest and support of musicologists. In December, we put out our call for expressions of interest in applying for a mini-project and received around thirty responses. This resulted in around forty people accepting invitations to the networking meeting; applications are expected to be for collaborative projects between musicologists and technologists and some of the expressions of interest were represented by two people at the meeting.
The programme for the day included some talks from the project team in the morning followed by introductions of the invitees, each making a 45 second statement about their proposed projects.
Alan opened the meeting with an introduction to the mini-projects, re-iterating the purpose and the requirements.
Tim then followed this with a general overview of Transforming Musicology, including recounting some of its history. Although Music Information Retrieval has a background in musicology and particularly music library science, the ISMIR community has shifted the focus since around 2000 more towards audio engineering and all the interesting problems that work has to deal with. Tim and his colleagues have consequently been promoting the potential of MIR to musicologists through a series of projects which he briefly outlined: OMRAS, MeTAMuSE, OMRAS2, and Purcell Plus.
Tim described the three principal components of Transforming Musicology: musicology, drawing on deep human understanding of music; music information retrieval, allowing computational searching and analysis of music; and the Semantic Web, allowing global exchange of cultrual concepts. For each of these components Tim identified the potential communities of interest:
- Musicology: musicologsts, enthusiasts, music-hobbyists;
- Music information retreival: computer scientists, audio engineers, data scientists, music industries;
- Semantic Web: other disciplines, global public.
The following three talks were given by Kevin Page, Christophe Rhodes and me, and each covered one of these areas.
Kevin gave an introduction to Web architecture, the Semantic Web, and Linked Data, making an argument for harnessing these technologies to make scholarly activities scale: rather than just publishing on the Web, scholars should be attempting to build richly interlinked resources taking advantage of Web architecture's resilience, the Semantic Web's affordances for sharing meaning, and Linked Data's principles for identifying and associating intellectual resources. Kevin gave an example of some exploratory work which made use of these techniques in answering the question, "How country is my country?" It used online data giving the country of residence (Geonames) of music artists (DBTune), online sources for music (MusicBrainz, Jamendo) by those artists, carried out a genre classification of the music, and published the results online—including linking them back to the source music and artist data (Linkedmusic.org). Once that genre information was online, it was possible to find the proportion of 'country' artists from any given country using all Web-accessible data. Transforming Musicology will be exploring such Web publication of data more deeply and successful mini-projects will be expected to employ these techniques.
Christophe gave an overview of some of the technologies available for music informatics, dividing them into the categories of: feature extraction; search; classification and visualisation; and online data access. He described extracting features from audio to find things such as pitches and then exploring those features using the Sonic Visualiser tool. In keeping with the Semantic Web theme, Christophe also described using Sonic Annotator in order to publish feature data online. As well as audio, Christophe also showed the
music21 tool for working with features from notation (or symbolic) music data. For musical search, Christophe outlined techniques ranging from Denys Parsons' code for "sorting" musical incipits (thus easing search) by contour, through other forms of musical summarisation, to the AudioDB audio search engine. Data mining is the umbrella term used for a variety of semi-automated data classification methods providing ways of exploring large collections of data. Christophe showed the WEKA and jMIR toolkits as examples for music data mining. He mentioned a number of APIs which make musical data available online, drawing on Paul Lamere's recent list. Finally, Christophe gave some pointers for further places to look including: http://code.soundsoftware.ac.uk/ and http://ismir.net/.
The final talk of the morning was given by me. I outlined some examples of recent work in "digital musicology". The studies cited were:
- Ian Knopke and Frauke Jürgensen (2012); "Symbolic Data Mining in Musicology". In Music Data Mining, Boca Raton: CRC Press.
- Elena Alessandri (2014); "Reviewing critical practice: An analysis of Gramophone’s reviews of Beethoven’s piano sonatas, 1923-2010". Musicae Scientiae (18:1).
- Nicholas Cook (2007); See publications of the Centre for the History and Analysis of Recorded Music project.
- Dmitri Tymoczko (ca. 2008); See slides on Functional Harmony.
- Reinhard Kopiez (2009); "Clara Schumann's collection of playbills: A historiometric analysis of life-span development, mobility, and repertoire canonization". Poetics (37).
- Peter van Kranenberg (2007); "On Measuring Musical Style — The Case of Some Disputed Organ Fugues in the J. S. Bach (BWV) Catalogue". Published online and in Tonal Theory for the Digital Age, Stanford, CA: Stanford University.
- Anja Volk (2013); "A corpus-based study on ragtime syncopation". In Proceedings of the International Society for Music Information Retrieval, Curitiba, 2013.
Many of these studies required a significant data acquisition and preparation task; nine-month mini-projects will need to consider carefully the data available to them. This spread of studies also demonstrated a variety of data types, both musical and extra-musical.
The afternoon was taken up by smaller break-out discussions on four different themes: Text processing and ethnography; Audio analysis; Mobile and web technologies; Pattern discovery and analysis of symbolic data. I was chairing the Text processing and ethnography discussion. We covered topics including data acquisition and especially the problems of marking up data and of dealing with optical recognition (both in text and music). We also talked about music metadata and the problems around reliably identifying musical works. Related to this we discussed finding and developing ontologies for publishing research outputs within the Semantic Web.
Following this was a short plenary in which Alan reviewed the requirements for mini-projects and each of the break-out sessions gave a brief report. The questions of data acquisition and publication proved to be quite important discussion points here too.
Now it's up to the interested parties at this meeting to team up and put together their applications. We're very excited to see what comes in. As the projects progress we will ask the award holders to report their progress on this blog, and hopefully also encourage some more real-time reporting on Twitter at #TMusMiniProjects. Do follow along!
EDIT: 2014-02-26 13:23: Added links to slides.