| Competition | Previous Entries & Ideas | Digital Collections | TOCs | FAQs | Judging | Resources and Tools | Submit Entry | Events |

Julia Craig-McFeely

Music Decoder

Abstract

A system that will recognise music characters in medieval manuscripts and render the results searchable to melody searching. It uses Gamera optical character recognition (working pilot: http://ddmal.music.mcgill.ca/liber ). A bonus is that once the shapes are classified it is possible to search for scribal characteristics, allowing a user to find similar scribes to a given exemplar.

Assessment Criteria

The research question / problem you are trying to answer*

Please focus on the clarity and quality of the research question / problem posed:

Music is not currently searchable from manuscripts unless it is first input into a search database. The Gamera system searches MS images directly and can allow searching on exact melody, or transposed or outline similarities. Tried and tested with a printed and a MS chant source (the latter including a mashup with a chant database to show information about each page: http://ddmal.music.mcgill.ca/salzinnes ): we want to do this with medieval music, which is rich in cross-repertory borrowings that cannot be located any other way except by chance. We have about 40,000 images already, and would like to extend also into early modern music using the BL content.

Please explain the ways your idea will showcase British Library digital collections*

Please ensure you include details of British Library digital collections you are showcasing (you may use several collections if you wish), a sample can be found at http://labs.bl.uk/Digital+Collections

BL music manuscripts are some of the most important and wide-ranging in the world. Creating searchable access to their content would vastly enhance their impact and usefulness in research.


Please detail the approach(es) / method(s) you are going to use to implement your idea, detailing clearly the research methods / techniques / processes involved*

Indicate and describe any research methods / processes / techniques and approaches you are going to use, e.g. text mining, visualisations, statistical analysis etc.

Gamera OCR software is specifically developed for music characters, but has been tested also on Greek and other non-standard alphabets of non-cursive writing. The pilots are in place, but development time, and more content is needed to improve its output. The software is adaptive, so the output has to be corrected, ideally by graduates, who are faster and more accurate than non-experts, but a well-trained non-expert could equally do this job as it requires little more than good shape recognition. Each time the output is corrected, the OCR becomes more accurate.

Please provide evidence of how you / your team have the skills, knowledge and expertise to successfully carry out the project by working with the Labs team*

E.g. work you may have done, publications, a list with dates and links (if you have them)

We would need to work with the Gamera development team at McGill University, Montreal.

Digital document image retrieval using optical music recognition
http://ismir2012.ismir.net/event/papers/577-ismir-2012.pdf

Creating a large-scale searchable digital collection from printed music materials
http://www2012.wwwconference.org/proceedings/companion/p903.pdf

Diva: A web-based high-resolution digital document viewer
http://link.springer.com/chapter/10.1007%2F978-3-642-33290-6_51

Searching the Liber Usualis: Using CouchDB and ElasticSearch to query graphical musical documents
http://ismir2011.ismir.net/latebreaking/LB-10.pdf

This one is a bit out of date as they are now using Solr instead of CouchDB and ElasticSearch, but the method is exactly the same.

Please provide evidence of how you think your idea is achievable on a technical, curatorial and legal basis*

Indicate the technical, curatorial and legal aspects of the idea (you may want to check with Labs team before submitting your idea first).

Using images from the BL collection would manage licensing issues. The additional content provided by DIAMM is covered by a rights licence that encompasses research and development using the image content, but each source would be double-checked with the document owner. The DIAMM database is available to provide manuscript metadata from which the content can be searched and pre-classified into notation types.

Evidence of achievability is to be found in the two pilots:

http://ddmal.music.mcgill.ca/liber (searchability)

http://ddmal.music.mcgill.ca/salzinnes (content mashup between the Diva viewer and a disparate database)

Please provide a brief plan of how you will implement your project idea by working with the Labs team*

You will be given the opportunity to work on your winning project idea between July 6th - October 31st 2013

Technical expertise from the Labs team would be essential in supporting the creation of user front end and delivery system, in working on usability, and also in establishing a group of users to correct the output from Gamera and feed it back to the software.