List of Participants

Musicological team

Technical team

Timetable

25 June 2003

Workshop Report

Introduction and aims of the project: John Rink

John Rink explained in detail the aims of this pilot project, which is funded by the Andrew W. Mellon Foundation to explore the key issues in the creation of an online variorum edition of the works of Chopin, but which will also elucidate the problems and propose some solutions to the editing of complex textual traditions of the works of other major composers. The project will be known as the OCVE project - Online Chopin Variorum Edition.

The recent work (funded by the Leverhulme Trust) done by Rink and Christophe Grabowski in the preparation of their annotated catalogue has identified 4500 distinct versions of Chopin first editions alone in some 50 libraries worldwide. The Annotated Catalogue provides a detailed description of each first edition and later impressions thereof, facilitating identification of the sources used to prepare them, providing information about the number and chronology of impressions, and paving the way for analysis of the interrelationships between multiple editions.

This vast complication came about because inadequate copyright protection between the principal European countries during the early nineteenth century compelled Chopin to employ different publishers in France, Germany, and England, thus giving rise to three "first editions" of most pieces. Each is unique, as a result of his idiosyncratic editorial methods and ongoing compositional revisions. Further differences arose from the interventions of house editors and proof-readers in successive impressions which until recently have simply been regarded as "first editions" - an error of judgment that has undermined much Chopin scholarship. In particular, German and English publishers tended to tinker incessantly and sometimes intrusively. Scores which can plausibly be described as 'first editions' despite such editorial and other changes continued to be produced until 1879 in Germany and well into the 1890s in England.

There may be over 10 different versions of any one Chopin first edition, and to trace the full publication history of individual first editions requires detailed and labour-intensive manual inspection and comparison of the relevant materials. The versions vary in quality of image as the same plates were used over and over or because of the loss routinely incurred upon lithographic transfer, and the paper was also subject to shrinkage, so that the same plate might produce images of different sizes over time. The publication history of the different versions has been thoroughly investigated within the the Leverhulme-funded project, but ongoing refinement will be required as part of the OCVE research.

Rink also explained that major funding has been requested from the Arts and Humanities Research Board's Resource Enhancement programme in order to create an online resource uniting all of Chopin's first editions in an unprecedented virtual collection. The archive will be drawn from four partner libraries (Bibliothèque Nationale de France, Bodleian Library, British Library, and Chopin Society Warsaw) and eighteen other institutions, totalling 4,345 digital images of Chopin's music. The full score of each first impression will appear along with analytical commentary on particularly significant textual features. In addition, there will be excerpts from the Annotated Catalogue. Innovative methodologies for complex textual interlinking and web delivery of this material will be devised, using advanced imaging techniques allied with relevant open standards for metadata and interface design. The application to the AHRB was submitted in May 2003, and their decision will be announced in November 2003. If unsuccessful, the project team (Rink, Marilyn Deegan and Harold Short) will actively pursue other potential sources of funding for this project. While success with this will greatly enhance OCVE, each project has been conceived and would proceed independently.

There are however similarities between the potential AHRB-funded project and the Chopin Early Editions project at the University of Chicago Library (http://chopin.lib.uchicago.edu/). These are being explored by the project team. The OCVE pilot project will benefit from the Chicago digital image archive in that individual images can be borrowed and scanning definitions and other benchmark references either emulated or developed further.

Although numerous variorum projects exist in the field of textual studies, many of which exploit the latest technologies with regard to image manipulation and collation/cross-referencing across discrete filiation chains, musicology has not yet exploited the application of such technologies to complicated source networks such as those pertaining to Chopin as well as to Bach, Mozart, Beethoven, and other composers. The aim of the current pilot project is to capitalize upon emerging technical capacities for text/image comparison and new music-recognition technologies that allow unprecedented manipulation and comparison of finely-grained musical elements. An online variorum edition of Chopin will be created as a prototype for later, more ambitious research on Chopin and hopefully also on other composers. Its primary scholarly goal is to facilitate and enhance comparative analysis of three categories of source material: manuscripts (sketches, autographs, scribal copies, glosses in student copies, etc.); first impressions of the first editions; and later impressions of the first editions (i.e., those pages of the first editions containing variants, whether attributable to the composer or to others involved in the editorial process). Three types of image manipulation will be investigated by the project:

It should be noted that the variorum edition need not stop at first editions: in principle it could include all the Chopin editions that have ever been produced, and it could also incorporate sound recordings for comparison. For use of any of these materials, we would need to deal with issues of permissions and copyright (as indeed for the manuscript and first-edition sources to be dealt with in the pilot project and in the first phase of the OCVE).

Methods

The pilot project will develop in particular the first two types of image manipulation referred to above. The third - combination/interpolation - will not feature in this project itself, although attention will be paid to eventual possibilities as well as potential problems.

Superimposition

It must be stressed that the project will be using images of the musical works themselves, not transcriptions, and one of the questions we will be asking is how to automate the process of comparison given that discrepant elements between different versions within a source pool are often very minor and thus difficult to discern even upon careful inspection. We will therefore investigate ways of visually superimposing printed sources within defined, discrete filiation chains in order to reveal the variants between them, which would be highlighted in colour and/or by other means. The requisite technology is already advanced in non-musical areas, although it will require modification. It might be possible to adapt Aruspix, a technique developed in Switzerland (see www.unige.ch/lettres/armus/music/devrech/aruspix/index.en.htm), the potential application of which will be investigated.

Juxtaposition/comparison

For the pieces chosen for the pilot project (to include Chopin's Preludes Op. 28 Nos. 4 & 20 as well as one large-scale piece, the Barcarolle Op. 60, extracts of which were shown during this presentation), the entire source will be digitized and separated out into bars. At the first stage of the project, the bar will be the smallest piece of information (not the note) because we are looking at the physical attributes of the sources. All bars will be made available, but we will be investigating some way of highlighting those which are particularly significant, given that a user (whether scholar or performer) may want to pull up quickly the variant bars in different versions. We will also investigate automating this process, but at the first stage we will need to do this as annotations. We might also want to build up user annotations to significant bars over time; thought will be given to the logistics and the scholarly implications of this.

As implied above, there will then be two modes of juxtaposition and comparison:

In order to offer some of these features, we need to decide how we represent the works via sigla, code (e.g. as in the Annotated Catalogue) and/or in terms of other codicological information.

Such an approach will allow the users to investigate and to reconstruct for themselves the creative history of a composer's work from a number of different perspectives. It was pointed out that this might mean that some users may create bad histories, but the goal of this project is not to police historical investigation or prescribe the 'good' or 'bad' qualities thereof - rather, to make possible for the first time, or at least to facilitate, such investigation.

General discussion

The participants of the workshop were excited at the possibilities offered here, and asked many questions and made a number of useful suggestions. There was debate about who the audience for this might be, and agreement that availability online would draw in new audiences. It was suggested that to give the best information for users of the resource, we need to understand a composer's practice in working with publishers. There was some discussion about the degree to which we would have annotations, commentary or an explanatory framework: if a user wants to call up all the variants, can we make a database of editorial commentary to present with the editions? That had not been part of the original proposal, but perhaps tools could be provided for these to be added later, along with the (admittedly considerable) scholarly information that would have to be compiled. It was noted that there are certain attractions to presenting the material 'neutrally', i.e. without the particular scholarly steer that OCVE researchers would inevitably impose by means of such editorial commentary; the debate launched here will need to be taken forward to weigh up the relative merits of the two approaches. Nevertheless, all agreed that we will need to create a framework within which the user can make informed choices about the sources on the basis of some sort of academic/scholarly apparatus. It was also agreed that the resource would depend greatly on having a repository of good data, and questions about quality of the originals we will work with will need to be addressed. Other issues raised during the general discussion included copyright and permissions, and access to sources (e.g. whether or not libraries and other institutions as well as private owners would allow us to use them).

One possibility noted above is whether the variorum edition could be used by a performer to create their own performance score, by means of the third type of image manipulation that we envisage: i.e. combination/interpolation of elements from disparate sources in tailormade collations. Not only must certain thorny aesthetic questions be addressed here, but there are design implications in terms of providing a screen that can be used at the piano keyboard, all of which will need to be considered in the wake of this pilot project.

Recently developed web technologies for image/text display and interlinking

Harold Short first of all set the technical context for the work of the variorum Chopin project. He introduced the work of the Centre for Computing in the Humanities (CCH) at KCL which is involved in projects in all discipline areas in the humanities. See http://www.kcl.ac.uk/humanities/cch/projects.html for details of these projects. His work at CCH has shown that tools developed for one area are often needed in other areas, and that some of the underlying paradigms can illuminate areas that they weren't constructed for. He suggested that one of the consequences of the accessibility of technologies is that there are now very few scholarly research projects which do not use technology. He showed the example of the Corpus Vitrearum Medii Aevi (CVMA) digitization project at the Courtauld Institute (http://maple.cc.kcl.ac.uk/ps/cvma/). CVMA was established in 1949 to identify and describe medieval stained glass, and in the past, the main output has been high quality conventional publication, with all its limitations. Now CVMA are looking at new modes of electronic publication as well as capturing high quality images of medieval stained glass, much of which derives from Picture Archive of the National Monuments Record (NMR). CCH and CVMA have experimented with republishing existing CVMA volumes in digital form because this way they can represent more images at high quality for scholarly use. Texts are converted to XML, images are scanned and edited, information goes into a database to link text and images. CVMA started as an art history project but the scholarly information is valuable to a much wider audience. The scholarly commentaries here could serve as a model for the ones that participants in the previous discussion asked for in OCVE; likewise the navigational possibilities demonstrated by Short with regard to the ground plan of a medieval cathedral.

Another multi-disciplinary/multi-technology scholarly editing project with some similar aims to OCVE is the William Blake Archive being produced by scholars at three institutions in North America (http://www.blakearchive.org). This is a hypermedia archive which is presenting Blake's writings with their visual content: this is key to the understanding of Blake's poetry. Tools have been added to allow manipulation of images, and it is possible to compare plates in two different editions, something it might be useful for OCVE to investigate. Additional information is available in the form of editors' notes (again, this kind of scholarly support having a potential parallel in OCVE), and the Inote image annotation tool allows the editors to put commentary on various parts of the image (e.g. in pop-up boxes of potential relevance to OCVE).

In discussion, reference was also made to such examples as the Gutenberg Bible and Göttingen Model Book projects (http://www.gutenbergdigital.de/), and online illustrations were provided.

Marilyn Deegan gave an overview of some other editing projects, musical and textual, echoing Short's point that certain paradigms have ramifications beyond their original contexts/purposes and that 'enabling spaces' are now sought in contrast to early problems of transferability.

It was noted that visits by the OCVE team to at least some of these projects would be useful - e.g. Beethoven, Mozart and/or Bach; alternatively, relevant experts could be invited to London for discussion in a colloquium/small conference.

Julia Craig-McFeely then discussed the Digital Image Archive of Medieval Music (DIAMM) project (http://www.diamm.ac.uk), focusing on topics of particular relevance for the OCVE project. She remarked upon the massive changes from 1996 in terms of web usage which projects can now take advantage of. The main aim of DIAMM is to obtain, archive and, where necessary, enhance digital images of European sources of medieval polyphonic music, digital encoding having been chosen because of its longevity, potential for manipulation and restoration, and convenience and potential availability. The project has been in existence since 1998, and Craig-McFeely remarked that libraries and archives had made a huge leap of faith in allowing the digital photography to happen. Some digital images are ordered from institutions which can provide them; in other cases, DIAMM staff go into the archive and take the images, subject to strictly enforced legal agreements (see below) - a procedure which might or might not be available to OCVE researchers. Where necessary, tools such as Adobe Photoshop are used for the restoration of images from damaged originals: this has been much easier and yielded images of higher quality than originally expected. The potential application to OCVE is obvious, in that many print sources as well as manuscripts are heavily stained, foxed or otherwise visually compromised and would thus benefit from digital clean-up.

DIAMM have attempted to impose its own (high) standards on archives, which is not always easy (and here again the implications for OCVE are clear in terms of quality control and specification/definition of scanning standards). Metadata is not added to the images, other than detailed technical and locational information. For content description, DIAMM relies upon existing catalogue records - though Craig-McFeely pointed out that to date some 30% of catalogue records have contained erroneous information. Scanning is done using a camera back that can deliver 144M pixels, giving file sizes of 350Mb. This means that DIAMM is able to reproduce even large manuscript images at real size. DIAMM has huge storage needs: some of the restored images are 5 x as large as the original, giving sizes of c. 2 Gb each. Issues of storage capacity and delivery capabilities (see below) will of course need to be addressed in OCVE.

Craig-McFeely suggested that there might be a danger in giving too many tools to users because of the possibility that they could come up with bad results if not trained in their use. She also pointed to the difficulty of deciding what is reproduction and what is editing. She suggested that many scholars might not be prepared to put their scholarship on the web as part of projects like DIAMM and OCVE, something that both projects have been considering as part of commentary, annotation, etc. DIAMM was conceived of from the beginning as an archive, NOT a delivery system: it was set up to rescue fragments that were in danger of disappearing, and indeed has unearthed other important fragments of medieval music. However, the audience for DIAMM is very keen that there should be a delivery system, and developments in web technologies since the project started have made this more feasible. DIAMM was accordingly awarded a planning grant by the Andrew W. Mellon Foundation for the investigation of a delivery system. The tools to be investigated will have some overlap with OCVE and the technical developments for both are being led by CCH.

Issues in common between OCVE and DIAMM thus include legal matters (DIAMM's development of licences & detailed legal documentation could be directly applicable), storage/delivery implications, digital imaging standards and modes of capture, implications of making 'editorial' changes to images, the creation of a Scholar's Workbench, print disabling (e.g. PDF rather than JPG), risk of image impairment through compression, the need to combat colour infidelity/inconsistency across users' different display screens (reference was made to the Bach digital thumbnails, which show colourscales and greyscales for calibration purposes), and pop-up content (e.g. page and bar nos. in OCVE).

Tim Crawford presented a wide range of relevant projects in musicological research, which are summarized on his (temporary) web page at http://www.soi.city.ac.uk/~timc/chopin/.

John Bradley pulled many of the strands together that were discussed throughout the day with the presentation of a proof-of-concept for OCVE using materials available online from the Chopin Early Editions project in Chicago (specifically the Prelude in C-sharp minor Op. 45 as published by Mechetti and Schlesinger, i.e. the Austrian and French first editions respectively). He first separated the music into bars (using the Map Edit software). He then drew boxes around the bars, attached a label to each one, mapped the bar locations on the page, and created a bar location database which could be used to assemble collections of bars in different versions of the sources and to create a complex system of navigation. In doing this he found a number of problems that would need to be resolved, including the key one of overlapping hierarchies: bars, slurring, stave, etc. He remarked that we are trying to capture two different aspects of the sources: image - what's on the page? music - what is the content? He then experimented with some different ways of presenting the content in a web environment. (In this respect comparison with the Chicago project will be particularly useful, in that similar questions of content definition and mark-up had to be confronted there.) He proposed that we could attach notes to parts of the text, with discrepancies marked up with popup annotations. These could be used to help analyse the sources. He suggested that it would be possible to categorize annotations so that they could be more useful, for instance so that the user could search for different kinds of annotations (e.g. with regard to changes made by Chopin himself, as opposed to house editors). He pointed out the need for a tightly controlled vocabulary and syntax for adding annotations (which the scholarly researchers within the full-scale OCVE project would need to attend to), but suggested that there could also be some free text. He then looked at some aspects of a possible Scholar's Workstation for OCVE (that could be generalized to other projects - cf. the DIAMM discussion above). This would present a database of sources to the scholar, with some editorial materials as part of the sources. The scholar would be able to add his/her own materials which could be stored locally in a personal workspace or 'published' to the edition, clearly marked as being produced by that particular scholar (perhaps going first of all into a 'quarantine' area for peer review). The whole could constitute a series of online collaborative editions with signed up contributors, editors, etc. to which access could be regulated as necessary. The Workstation could also have a tool for creating 'pathways' (cf. Perseus Project) which could be used for giving presentations, teaching, sending to other scholars for discussion points.

Breakout group: technical

Breakout group: content