ECOLM III: opening historical music resources to the world's on-line researchers

Lead Research Organisation: Goldsmiths University of London
Department Name: Computing Department


ECOLM III will semi-automatically build a computer-readable corpus of around 2,000 pieces of vocal and lute music from the 16th century. It draws on the new Early Music Online (EMO) collection of 300 digitised books of 16th-c music from the BL, and will use state-of-the-art optical recognition methods, presenting the results both in scholarship and in professional performances. The resulting encoded scores, of both vocal and lute music, can be manipulated in many ways: audio playback, reformatting, printing, editing and conversion into other formats. They will be made freely available to all, amateur and professional scholars and musicians alike, via the ECOLM web-site.
For musicology to benefit from the technological advances currently transforming text-based scholarship, we need machine-readable encodings of musical scores. At present, a few resources provide suitable high-quality encodings of parts of the central canon of classical music, but very few are done directly from historical sources rather than from out-of-print standard editions. The Electronic Corpus of Lute Music (ECOLM), originally set up with AHRB funding by Tim Crawford in 1999, has consistently concentrated on encoding complete contents of sources of lute music to high scholarly standards.
Lute music uses a special form of notation called tablature, which only players of the instrument can read with ease. Using optical character recognition methods optimised to work with 16th-century printed music (partly as a result of pioneering trials carried out within ECOLM II), encodings can now be done semi-automatically using optical tablature recognition (OTR) in the case of lute music or optical music recognition (OMR) for the printed part-books of vocal music (in more normal notation).
For correction of errors in OTR and OMR (caused by problems with 16th-c printing or damaged sources) we will engage the help of the people who know these forms of notation best. For lute music, these are members of a world-wide community of amateur lute players who are both highly active online in sharing images and used to reading tablature; for vocal music we will enlist the help of musicologists and other musicians, amateur and professional. The corrections made, using web-based editing software we have developed, will feed back into the recognition process of each system, gradually improving accuracy until eventually almost no human intervention is required.
Our workflow will be designed to ensure systematic corpus growth, prioritising vocal collections with works also in the EMO corpus as lute arrangements. We expect to encode semi-automatically around 1,000 lute pieces and about the same number of vocal works (mainly French chansons), with significant overlap of repertory. We shall carry out preliminary musicological investigations on topics such as: (a) relationships between lute arrangements and their vocal models, (b) identification of characteristic melodic and harmonic patterns (e.g. cadence figures) in both repertories, and (c) the way in which these changed over the century. These form the basis for future investigations of the growing corpus.
The project will end with a public workshop on the project and its outputs at the BL, and also a concert at the prestigious nearby venue, King's Place, promoted jointly with the BL and the UK Lute Society, in which fruits of the project will be made manifest in professional performances of the highest international standard.
ECOLM III takes advantage of new opportunities arising since the conception of the second phase of AHRB/C funding for ECOLM II. With our partners, the BL Music Dept and the UK Lute Society, we shall work closely with the EMO resource to maximise mutual benefit and to ensure its rapid uptake and impact not only by academic musicologists and professional performers but also by a world-wide community of amateur singers, lute-players and other musicians.

Planned Impact

By involving our partners, the British Library and the UK Lute Society, this new and innovative phase of the Electronic Corpus of Lute Music (ECOLM) will have impact within and outside the academic community, involving professional and amateur musicians across the world.
The key novel and exciting aspect of ECOLM III is the interaction between two communities: professional musicologists and a larger, non-academic, world-wide community of expert interest - those who play the lute from its original tablature notation purely for pleasure. The UK Lute Society has around 1000 members world-wide, and the American Lute Society has 700; allowing for overlap and estimating numbers for the French, Belgian, Dutch, German, Spanish societies, we reach a figure of some 3-4,000 world-wide. If only 5% took part, 150-200 tablature experts will contribute to the corpus's rapid growth.
A recent JISC award to the BL and RISM UK for the Early Music Online project (EMO) makes available 300 printed books of 16th-century music from the BL as digital images. These allow ECOLM III, in collaboration with EMO, to present about 2000 pieces of vocal and lute music in computer-readable form for the first time, using state-of-the-art OCR techniques with error-corrections done by our world-wide community. These high-quality encodings represent a step-change in musicology, supporting rather than threatening traditional values in the discipline. The enhanced online facilities of ECOLM will open this music to anyone in the world with a standard web-browser.
ECOLM III is unique in that it will produce a significant corpus of vocal and lute music in machine-readable form with minimal human intervention. This is unprecedented with musical source-material, and it is to some extent analogous with text-OCR-based projects, such as Google Books, or JSTOR, which are rapidly changing academic practice. As well as being ground-breaking in terms of music data-curation, it will also have a major impact on musicology, in that an important section of the repertory of renaissance music will become available for deep study quickly and at very low cost.
Recent advances in web technology allow musicians of all kinds to share materials freely and instantly, leading to a great surge in informal online publication. Despite well-recognised problems of quality control associated with self-publication, the fact is that musicians increasingly expect to be able to find the music that interests them online, in an instant. By contrast with the arbitrary choices of non-scholarly online publishers, ECOLM takes a systematic and exhaustive attitude to the encoding of music by completely and as accurately as possible representing the original source(s). This scholarly attitude will give significant world-wide impact for ECOLM as the corpus grows.
ECOLM III holds interest for both professional and amateur musicians and for Humanities Computing in general; a public Workshop, jointly organised with the UK Lute Society at the BL, will launch a post-project phase of more encoding and also promote the general aims of ECOLM and related projects, demonstrating new possibilities for scholarly study. The scope will also be widened to include those whose primary interest is in the field of vocal music. We will thus promote a wider range of interactions between scholars and practical musicians as well as drawing interest to a number of issues surrounding historical sources and texts and modern performance.
Finally, a concert at the highly-prestigious venue, King's Place, jointly promoted with the BL Music Dept and drawing on our encoded musical corpus, will bring together scholars and professional musicians to realise the project results in performance for a wider public. This will be well publicised by us and the BL, and - subject to negotiation and further financing - recorded and/or broadcast, so ECOLM III will culminate in an entertaining yet scholarly musical activity with maximum public impact.
Description We have carried out research into ways in which historical music resources on the web, in our case the Early Music Online (EMO) collection of images at the British Library (hosted by Royal Holloway University of London), can be exploited by encoding them as musical scores. We worked with two specialised systems for Optical Music Recognition (OMR) together with their developers (consultants to the project): the first, Aruspix, is for printed vocal and ensemble music of the 16th century; the second, Gamera OTR, is for lute music printed in tablature.

EMO contains around 300 books of music printed before 1600, representing several thousand pieces of music in total. Once the images have been preprocessed to ensure the best possible recognition performance (the BL gave us access to their own archival high-resolution TIFFs for the purpose), they can be run through the OMR software. However, there are various problems which need to be overcome before useful musical data (in the form of editable music-notation) can be extracted. The first of these is inevitable: some pages cannot be recognised because the images ar simply not suitable - either of poor quality (in terms of photography), in which case they could be re-shot, or of condition, such as the many cases where there is 'show-through' from a verso due to the thin paper on which the music is printed. We produced a paper on this problem for the 2013 ISMIR conference.
The second problem is that of metadata. The BL online catalogue for EMO items was redone when the books were digitised (from high-quality archival microfilms), and they give complete listings of the musical contents of each book. Unfortunately, the standard library format used (MARC) does not allow detailed specification of works within a volume, or how a work is represented and 'distributed' in multiple part-books. For this reason the contents-lists exist within large, informally-formatted text fields which vary in their consistency and accuracy. We were forced to perform a laborious process of data-preparation in order to ingest the information about the music in each book into a relational database for EMO, and to reconcile this data with our existing ECOLM database.
The third problem, closely related with the second, is that we need to be able to associate musical works in the collection with precise page-locations in the images, and vice-versa. This is further complicated by the fact that, just like manuscripts, these early music-prints are often somewhat haphazard in their organisation. A piece in a single partbook may have two or more voice-parts displayed on an opening; the voice-part for a piece may begin at any point on the page, and end at any page - sometimes the same page, but often at least a page later, or even on an earlier page with a 'direct' graphic symbol to indicate this. Firstly, we needed to develop the database scheme to represent this diversity in data, but we felt it was necessary to have a more efficient entry method than a typical database form. Detailed manual annotation would be too time-consuming and error-prone, so we developed a drag-and-drop interface, using the latest web technology, which is intuitive to use and much quicker than manual entry.
Exploitation Route The technology developed, and the methods we used, have been furthered within the Transforming Musicology project, q.v.

It has also been a firm basis for developing a novel concept of 'playability' of music notated in tablature, which is further exploited in our Learn To Play project.

The network of amateur lute players who we recruited will be involved further in our Learn To Play project.
Sectors Creative Economy,Culture, Heritage, Museums and Collections

Description We put on a concert on 7 September 2012: "La Fleur des Chansons An evening of Renaissance favourites from the music collection of the British Library A showcase programme featuring some of the most popular music of the 16th century held by the British Library. Much of this enchanting repertory coexists in vocal part-books and in arrangements for the most popular instrument of the age, the lute, or for keyboard instruments such as the harpsichord or organ. A major portion of the vast riches of the British Library's holdings of renaissance music has recently become accessible for the first time to anyone with an Internet connection. This concert provides a glimpse of the treasures in the 300 16th- century printed music books so far digitised in Early Music Online (EMO) and the Electronic Corpus of Lute Music (ECOLM) which together include around 10,000 pieces of 16th-century music. Most printed books of music from the 16th century were intended for amateur performers to play in their homes, but they nevertheless preserve some of the great masterpieces of renaissance sacred music alongside secular fare of less serious intent. Some of the finest of that music was arranged with great skill by the most famous lute-players and published for amateur players to try at home. We also include a few examples of virtuoso instrumental music given in contemporary performance manuals for the benefit of competent players of other instruments aspiring to professional standards. The concert ranges from the poignant setting of the 'Pater Noster' (The Lord's Prayer) by Josquin des Pres, through to earthy dance music close to what would have been heard in the street, via some of the most popular and tuneful French chansons, including one about the power of love to conquer the toothache. We also include Italian part-songs and madrigals, some of which were published with words translated into English, thus forming the basis for the English madrigal. Performers The Brabant Ensemble (director Stephen Rice): Kate Ashby (soprano), Sarah Coatsworth (alto), Alastair Carey (tenor) and David Stuart (bass) with: Jacob Heringman (lute and renaissance guitar), Hector Sequera (lute), Emilia Benjamin (viols), William Hunt (viols), William Lyons ( recorder and renaissance flute), Steven Devine (harpsichord) Venue Hall 1, King's Place, 90 York Way, London N1 9AG Friday 7th September 2012 at 8 pm (doors open 7:45 pm)"
First Year Of Impact 2012
Sector Creative Economy,Digital/Communication/Information Technologies (including Software),Education,Culture, Heritage, Museums and Collections,Other
Impact Types Cultural,Societal