Mark Gotham reports of the motivations, history, and prospects for the OpenScore Lieder Corpus, a project for democratizing access to art song through encoding.
While many of us musicians and music lovers have a special place in our hearts for the tactile experience of printed sheet music, we are also rapidly recognizing the benefits to be had from their digital equivalents. These digital scores can take many forms, at several levels of sophistication, but even the humble PDF scan of an existing edition for performance from a digital device is a great boon to the jobbing musician. No more lugging around volumes of sheet music. No more worrying about whether the venue will have adequate lighting. No more cascades of music falling off the stand mid-performance.
Moreover, many of these PDF scores are now available instantly, for free, at the click of a button. The International Music Score Library Project (IMSLP) boasts a collection that is hundreds of thousands of scores strong and growing all the time, making it much larger than even most specialist music libraries, and very much more accessible. The scale and significance of this increase in access to sheet music is hard to overstate.
PDFs are great, and IMSLP’s scale is impressive, though viewing an image of a score on a screen is not so different from looking at the printed page. It is when those scores become truly computer-readable and -editable that they start to take on serious versatility. Many of us are familiar with at least some form of what it means to manipulate a score computationally through notation software like Finale, Sibelius, or the free and open-source equivalent, MuseScore.
Example of a score from OpenScore Lieder Corpus. At the top, one can see tools for transposing the score to any key, just one of the many benefits of these publicly shared encoded scores. Please click on image to enlarge.
Armed with this software and a score in the right format, you can execute a range of important tasks. One priority for singers, pianists, and teachers is the ability to transpose instantly to any key. So now we can add another item to the list: no more need to keep multiple copies of the same song in different keys (or spend time transferring markings between them). And that’s only one example task; the possibilities are endless as researchers like me, using computers to facilitate musical analysis, are starting to find out (but, that’s a story for another time).
Speaking of all this technology, is there not software that will simply do the encoding for us? Unfortunately, the answer is still a rather feeble ‘sort of.’ ‘Optical Music Recognition’ (OMR) software for converting from PDFs to encoded versions will one day make this task instantaneous, but it is not at that level yet. Moreover, the software is most likely to get to that level by incorporating machine learning, which in turn requires a large and reliable dataset from which to ‘learn’. Specifically, it needs score images (e.g. PDFs) paired up with a corresponding, encoded version. So apart from the many immediate benefits they offer, it may be that the main, long-term contribution of these publicly-shared score encodings is as training data for OMR.
Yet, despite these clear benefits, we tend to work with relatively few scores in those editable formats; often it is only the music that we personally have composed, arranged, or transcribed from other sources. Those transcriptions are only rarely shared, and even when they are, the licensing situation is often ambiguous at best, making it hard to know what constitutes fair use. Suitable legal protections should exist for copyrighted material, of course; for instance, it is entirely right that living artists should have proper controls over the compositions, recordings, editions, and other work they produce. Yet, the vast majority of our musical heritage has long passed out of copyright and into the public domain. Works composed, edited, and published in the nineteenth century are freely available to everyone. The creation of digitally encoded versions of those resources should not impose new barriers but rather support that free availability by enabling ever wider and more open-use cases.
In short, there is a need for better coordination of score encodings and at least clarity over (but preferably the complete removal of) license restrictions. This is the backdrop against which, in 2018, I sought and obtained a first seed grant to start building a collection of songs to be released in encoded formats and under a license which clearly, unambiguously permits the broadest possible usage. Two years later, the OpenScore Lieder Corpus is a large and growing collection of over 750 songs from a range of composers, all freely available online under the maximally permissive ‘CC0’ license. Anyone can use these songs for any purpose: we can perform from them, transpose them, play them back, make editions and arrangements, incorporate them into teaching materials and research projects, and more besides.
Beyond the opening up of functionality and the inherent democratization of access offered through online dissemination, this corpus promotes inclusivity in two other, important ways: in the representation of diverse historic composers, and in the inclusion of musical communities today not only in the use of the corpus but also in building it. These two groups both warrant special attention so let’s look at each, in turn, starting with historical composers.
This year’s 250th-anniversary celebrations of Beethoven will hardly make a difference to his public standing, whether they had run as planned, or as now, in their unexpectedly muted, COVID-era formats. The composers that really benefit from focused, sustained attention are the lesser-known figures who have been neglected for non-musical reasons, such as gender. This project very explicitly serves to support that diversification of the repertoire; indeed, we are working on songs partly because it is a repertoire in which more composers have been able to participate than, say, symphonies.
Unequal access to opportunities manifests in many and varied ways at every stage of a prospective career, and while we can do little more than acknowledge the absence of opportunities afforded to historical artists during their lives, we still can do a great deal now to cultivate, preserve, and disseminate those works which do survive. This serves to demonstrate that classical music is for all, and to encourage, embolden, and recognize the legitimacy of all artists today. In so doing, we expand, enhance, and enrich the repertoire.
As for the people building this corpus, we have adopted a structured, modestly incentivized crowd-sourcing approach that has enabled us to bring together a team of volunteers from around the world, many of whom are developing their own musical skills at the same time as serving this public mission. The act of copying out scores has been recognized as a way of learning music for hundreds of years and there is no reason why digital encoding should be fundamentally different. Thank you to the many people supporting the project as transcribers and reviews. Thanks especially to our associate at MuseScore, Peter Jonas, and our senior reviewers Dan Rootham and Mike Nelson for their tireless work not only in support of the corpus itself but also of its contributors as they progress along their own musical journeys.
We hope you find this to be a useful resource, complementing IMSLP and other provisions. If you have special requests for songs you’d like to see included, let us know. If you have a moment and would like to pitch in, then please do. We work in the MuseScore format but are happy to consider initial submissions of any kind. In any case, and even if you have no personal stake in any of this but are sold on the value of the project, please help spread the word as we seek to galvanize a more coordinated approach to the encoding and release of scores for the benefit of all.
August 21, 2020
More information can be found on Four Score and More.
