e-Science for Musicology Workshop Report
Funded under the AHRC/EPSRC/JISC e-Science in the Arts and Humanities Initiative, the e-Science for Musicology workshop held in Edinburgh in July 2009 brought together expert speakers in using computer technology in music scholarship and musicologist participants with little or no experience in such methods. During the course of the workshop, several techniques for dealing with music in computers were demonstrated and the participants were given the opportunity to comment on those techniques and discuss their wider implications for the discipline.
The Purcell Plus project, hosted at Goldsmiths College, University of London, is attempting to establish what an e-Science methodology for musicology may be like. What aspects of e-Science may be applicable to existing musicological practice? What new kinds of musicological practice may come about as a result of employing e-Science tools and methods? This investigation clearly requires that we engage practicing musicologists in the results and possibilities of e-Science techniques. The e-Science for Musicology workshop, therefore, was held as part of this process of engagement and, for the Purcell Plus project, of assessing the impact of e-Science on musicology. It took place at the National e-Science Institute in Edinburgh on 1 and 2 July 2009.
Speakers were invited to demonstrate tools and report on results of using data-rich and technology enhanced methods to aid musical scholarship. Participation was encouraged amongst musicologists of any stage in their career throughout the UK. Our aim was to foster engagement between musicologists who may have little or no prior experience of using technology in their research and those who are developing or using technologies for musical scholarship through demonstration and discussion.
Richard Lewis (Goldsmiths) opened the workshop with some thoughts on the nature of e-Science and on a scientific (or at least data-rich) approach to musical scholarship. Along with the traditional view of e-Science as distributed computing, distributed datasets, and collaboration over digital networks, he described a conception of e-Science (after Paul Wouters 2006) as three new modes of knowledge creation: computational discovery, comparative research, and digital library browsing.
Of these, digital library browsing is the practice most common in humanities research. Many humanist disciplines are involved with close reading of literature, and the availability of a good proportion of scholarly literature online is allowing scholars to do a lot more work from their desks. For musicologists, this ought to extend to access to musical materials (scores, recordings) in digital libraries. Several of the presentations in the workshop dealt with digital encoding of musical materials, publishing them online, and providing metadata describing them.
Comparative research involves generating new knowledge by analysing large quantities of data for statistical trends. This fits very well with scientific practice, but is less obviously applicable to humanistic research. In the case of musicology (and as was demonstrated in Nicolas Gold's presentation), there is scope for being able to analyse performance comparatively through analysis of numerous recordings of the same work.
Computational discovery concedes some of the task of hypothesis formation to the computer, through techniques such as automated inference of new facts from a knowledge base (a collection of facts and rules). Even in the sciences, this mode of knowledge creation has seen little application in research contexts. For humanists, though, the idea of computers being responsible for original theorising most likely seems quite jarring. However, one area of research that takes music as its subject matter does fit into this computational discovery mode of knowledge creation; computational cognitive modelling of music cognition attempts to form hypotheses about how the brain deals with musical stimuli and was the subject of Geraint Wiggins' talk. Casting the net a little wider, many applications of music information retrieval (MIR) techniques (the subject of both Stephen Downie's and Frans Wiering's talks) are concerned with attempting to assert new facts about collections of musical works by computational analysis.
Publishing Musical Data
One of the most well represented problems at the workshop was that of encoding and making available musical materials in digital form.
Ichiro Fujinaga (McGill) talked about the work going on in his lab, the Centre for Interdisciplinary Research in Music Media and Technology, in developing and improving technologies to aid the digitisation of musical materials and to provide centralised, Web-delivered access to those materials. His project, Distributed Digital Music Archives and Libraries (DDMAL), is attempting to develop technologies such as optical music recognition (OMR), to develop efficient workflows for digitising analogue recordings, and to specify and promote metadata standards for music libraries.
David Bretherton (Southampton) reported on the current status of the musicSpace project, which is applying the Semantic Web (for data description and sharing) and Web 2.0 (for advanced browser-based interfaces and user interaction) technologies developed in the mSpace project to research materials relevant to musicologists, including recordings from the British Library Sound Archive, articles from the Grove Dictionary of Music and Musicians, and music manuscript bibliographic records and incipits from RISM.
Yves Raimond's (BBC) main argument (after Tim Berners-Lee), in his talk on using Semantic Web techniques to describe musical knowledge, was that the Web is more about things than it is about documents. He described the ideal situation in which the Web is a graph of strongly interlinked URIs (names for things), and where all of the links bear relevant and useful semantics. For musical knowledge, he showed how "things" such as composers, works, and records and the relationships between them can be described using RDF and semantics based on his Music Ontology.
Doing Musicology with Data
As well as providing access to musical materials for scholarship, some presentations dealt with how musicology may be done using such data, and some of the problems that need to be addressed about doing so.
Daniel Müllensiefen's (Goldsmiths) talk (which was presented by Richard Lewis) reported some findings from the Modelling Music Memory and the Perception of Melodic Similarity (M4S) project. This work compared results of psychological tests of what constitutes a musically salient feature in a melody and computational models of such features. Some important concepts for doing musicology with computers were covered in this talk, particularly the idea of treating music as data and dealing with it syntactically in computers. The talk explained how this involves extracting features from the musical signal (in this case the signal was made of of MIDI events, but audio data is also commonly used) and how appropriate statistical analyses of these features can be applied to draw conclusions.
Daniel's talk also introduced some notions around the possibilities and limitations of corpus musicology, and what research questions become possible given access to a large collection of music and the means to investigate it as a whole. While some of the questions common in musicology cannot be answered by corpus analysis (aesthetic connotations, meaning and interpretation, and cultural and contextual explanations), other styles of questioning often neglected in musicology become possible (detecting general mechanisms working within a corpus, comparisons between corpora, and stylistic patterns and anomalies).
Frans Wiering (Utrecht) argued that the agenda of the MIR field has so far been set by the technologists and that its main research questions are basically technological in nature, covering such topics as searching for so-called musically similar audio examples from a database of audio data. His talk called for a re-configuring of this agenda towards the pursuit of musical meaning; a topic which, he argued, has become more central in recent musical scholarship. He described how the subjectivity of musical meaning stands it at a considerable distance from the syntactical view of music employed in MIR techniques, but he argued that MIR scholars ought to be considering how their methods may be applicable to questions of musical meaning and how findings in the musical meaning programme should influence the future direction of MIR research.
Modelling Music Cognition
Geraint Wiggins (Goldsmiths) described an application of computers in a very different branch of musically based scholarship: computational modelling of music cognition. His primary assumption is that music is a cognitive phenomenon, which he justified by demonstrating that a sonic stimulus does not carry any musical information but requires interpretation by a mind in order to become music. He described how one approach to study how the mind turns sound into music is to simulate models of them on computers. In particular, he reported on some of the results obtained using a model developed by Marcus Pearce called IDyOM (Information Dynamics of Music), which predicts the pitches in melodies using an information theoretic model. He went on to argue that the relationship between these results and musicology is that, while musicology can tell us what is important about music, such cognitive research can tell us why music is important.
Digital Editing of Music
Another well represented area of application of computers in music research is that of producing digital critical editions of musical works, corpora, manuscripts, or other written materials. This work is strongly related to that which goes on under the broader label of digital humanities.
Raffaele Viglianti's (King's College, London) main argument in his talk on the practice of digital editing of music was that all editorial practices involve the selection of what he called facts about the material being edited. In traditional published editions, this selection process is made by the editor and committed to paper before being handed on to the performer or scholar. Evidence of the selection and the possible alternatives selected from are only available in an appended critical commentary. Viglianti argued that the digital critical edition is significantly different from the published edition in that it allows the alternatives to be presented to the user of the edition orthogonally with each other and therefore allows the selection process to be deferred to the point of interpretation by the performer or analyst. This is made possible by employing a flexible representation method for the musical material and providing an interface that makes the selection of alternatives intuitive and easy.
Johannes Kepper (Paderborn) presented recent developments at the Edirom project. The project is developing software to aid the production of digital editions of music, including music notation editing and visualisation of critical comments orthogonally with the notation. He discussed some problems associated with representing score-based hierarchies in multiple versions, particularly problems such as how to resolve the conflict when bars (or measures, which form an intuitive structural hierarchy) are different in different sources.
Benjamin Bohl (Paderborn) demonstrated work from his masters thesis on producing an interactive edition of a treatise on composition by Francesco Geminiani (16871762) called Guida armonica o dizioniaro armonico. The treatise provided formulae for composition of harmonic structures by requiring the user to select between permissible continuations of each harmonic fragment from the catalogue of available fragments (numbering around 2000). Bohl's system encoded each of the harmonic fragments and the paths linking them, and provided an interface which, at each point in the compositional process, presented all the possible continuations from which the user can choose. He argued that, as well as being a digital edition of an interesting eighteenth century piece of work, his Web-delivered interface to it actually adds to and extends the usefulness of the information it contains.
Analysing Music with Computers
To complement the discussions of ways of encoding and publishing musical information, two of the presentations dealt with actually using digitised musical information for analysis.
Nicolas Gold (King's College, London) presented some results from the Analysing Motif in Performance project which came out of the Centre for the History and Analysis of Recorded Music (hosted at Royal Holloway). The project investigated the possibility of analysing performance through recordings. Their techniques involved providing visualisations of the complex data that can be derived from musical recordings, such as timing and dynamic information. As well as demonstrating an example of clustering timing data of numerous performances of a mazurka in order to show the relationships between different performance practices, Gold also reflected on some of the issues involved with disciplinary collaborations. He described how it was important to invest time in training and gaining experience in the discipline with which you are collaborating. He also described the importance of learning what he called the "norms and forms" of that discipline; finding out the kinds of questions that are valid to ask and the ways that results can be presented.
Philip Wheatland (Melbourne) demonstrated his MelodicMatch software, which is a tool for performing searches in musical score data. He showed examples of analyses of parody masses by Gombert in which he was able to show cadential placement and motivic development, both of which contribute significantly to the identity of a piece. The software allows the user to specify searches for melodic, intervalic, rhythmic, or lyric fragments and numerous other features. It also provides a visualisation in which matching portions of the music are highlighted in different colours and different searches can be displayed together. Wheatland emphasised the fact that the software is purely syntactical in the tasks it is able to perform and that the musical analysis must still be performed by the musicologist.
Sound and Visualisation
Richard Polfreman (Southampton) demonstrated his FrameWorks 3D software, which provides a visualisation tool for composers and sound designers. The software allows the user to deal with "clips" which can be fragments of audio, MIDI data, etc. and arrange them on "tracks". Relations may then be defined between clips and will be dynamically maintained by the software during subsequent manipulation of clips, tracks, and other relations. New classes of relations can be added to the software as plugins; it includes relations such as reverse by default. The clips, tracks, and relations are presented in a 3D interface that the user can manipulate with the mouse. As well as arguing that this interface reduces some of the complexity of dealing with and arranging ideas for composers, Polfreman also described its potential application as a music analytic tool by attempting to re-construct an existing piece of music using it and learning about the relationships employed by the composer in the process.
Subdiscipline or Methodology?
J. Stephen Downie (Illinois) described three institutions that support and promote the work of the music information retrieval field: the International Society for Music Information Retrieval (ISMIR), the Music Information Retrieval Evaluation eXchange (MIREX), and the Networked Environment for Music Analysis (NEMA). He argued that, while these institutions have fostered significant development in technology and encouraged and supported technologists, they have so far given little to musicological scholars who may be interested in using software in their research. As well as expressing the desire that musicologists should engage with the MIR community, Downie argued that a common language needs to be fostered between the two groups, and that a style of questioning and a research programme that engages both technologists and musicologists needs to be established.
The closing discussion of the workshop continued this theme, attempting to answer the question of whether e-Science approaches to dealing with music form a subdiscipline or whether they are just methods of another discipline (musicology or computer science, for example). It was generally agreed that, at the moment, such techniques were not widely applied in musicology and that computational musicology was still quite a separate discipline from the rest of musicology with not only its own methods and tools, but also its own research questions that aren't reflected in the research agenda of the other main branches of musicology.
However, it was also agreed that promotion of the methods of computational musicology amongst musical scholars may be a good way to encourage interest in a research programme that deals with music on the large scale and moves computational musicology from being a subdiscipline to being a key skill of all musicologists.
Wouters, Paul (2006). "What is the matter with e-Science? thinking aloud about informatisation in knowledge creation". The Pantaneto Forum 23 (July 2006). http://www.pantaneto.co.uk/issue23/wouters.htm
About the Author