 So next up we have a talk from Peter Jones is going to talk about open sourcing sheet music So as usual you have 15 minutes and the stage is yours Thank you Hi everyone I'm Peter and I'm talking about open score and it's our mission to open source sheet music so we last here back at Fosden 2017 where we announced open score to the world and We said that it was our mission was to basically do what what open street map did for maps or project good and I did for books We want to do for sheet music and specifically public domain sheet music So what does this involve for sheet music? So we're basically taking so initially paper scores that are in the public domain and And You it's a two-step process to create open source sheet music firstly you need to scan the music to create bitmaps But then you don't want to just stop with a PDF. You want something that you can Actually interact with we want to get musical source code So the next step is to convert it into an XML format so that you get a semantic score something that renders with nice crisp Crisp lines and that you can use to render Audio as well so you can actually listen to the score So the step one is covered by IMS LP, which is the the world's largest community of people up who look for public domain scores around the world in libraries and so on and Scan these scores and upload them to IMS so the IMS LP archive and The step two is covered by the Muse score community So this is taking those PDFs or FIM SLP and transcribing them to convert Excuse me to convert the PDFs to XML so Muse scores the the largest online community of sheet music creators and Of course, it's open source software under GPL version 2 So the idea of this is that we create digital scores something that you can you can actually listen to not like a PDF That you can edit so you can change notes or extract parts like instrumental parts Transpose the score. Maybe if you're a singer so that you can reach the notes And then once you've made your changes you want to be able to share those changes with other people without being restricted by copyright So we're trying to liberate music from copyright and liberate it from paper and So the liberation from Paper was covered by the XML conversion and liberation from copyright is done by publishing all of the Transcriptions under Creative Commons zero. So there's they're effectively fast-tracked into the public domain So there's complete freedom to to adapt them and share them so in preparation, so we announced the project FOSDEM 2017 and At that point we had run Two or one at that time pilot transcription to do the Tchaikovsky's six symphony And then shortly afterwards we did another to transcribe for his requiem so the method was that we would take an imslp PDF and We'd basically break it up into chunks of a few pages Send them out to music or users who would transcribe these pages and then we would check them and If they to see whether they were accurate with the original and if not then we'd request improvements send them back and Then once they were all finished we'd join them together to create a finished score and the result of this was that We we reckoned we could do it take about a month to fully transcribe a Symphony using this method, but of course you could run multiple in parallel to speed things up and The hope is that while it was initially Well myself doing the checking of all of these that we would be able to invite more experienced transcribers to take some of the review burden and Then it would be up to me to do a final review Before publishing it on the open score account so we then launched a Kickstarter campaign in June of 2017 and Called open score join the sheet music revolution and the pitch of the campaign was that We would liberate an initial 100 works using the tools that we currently had available at the time and So this was like a concrete goal That we had and then the sort of the more abstract aim of the project as well Was that by backing this project you'd also be helping enable it to grow in the future? So we would use this hundred works as like a trial at scale to try and establish the best procedures For doing the transcriptions to solve any scalability issues develop the required Automation tools to enable it to grow in the future to eventually liberate all public domain music So how it worked? was We appealed to Musescore users that if you Give us your time in transcribing the scores then we'll reward you with a Pro membership of Musescore's share a music sharing website and That the Kickstarter backers is that if you help us Back if you help the project by contributing money This will enable us to run the transcription effort and to make those scalability improvements and the the backers would get rewarded by they would have Their name would appear in the credits and they were perks that they could pay If they paid a certain amount of money, then they would get to choose an addition that we would actually digitize one of those hundred or they could And if they paid a bit more then they could actually write a dedication in the public edition So you could be the person to to have a dedication in the public version of Beethoven's fifth for example and As a bonus each of these hundred editions would get a Visualization made by the digital artist Nicholas through Joe Who had been? visualizing Muse score scores from Musescore.com before this and here's a visualization of his Of the Valdez four seasons that he created So he offered to create a visualization of each of the open score transcriptions so so we ran the campaign and That it lasted a month. We Exceeded our goal of 45,000 euros. We managed to raise 51,000 and over a thousand people back to the campaign and hundreds of people Musescore users told us that they were interested in transcribing So then the transcription work began straight away and The way this would work is I would open a transcription group on Musescore.com and Invite a bunch of transcribers and give them each sort of three to five pages of music They'd upload their transcriptions and then I would check them So the early challenges that we faced is the variety of pieces chosen by the Kickstarter backers these range from piano scores to entire symphonies And the the length of the pieces there are many scores with over More than like 10 instruments or 50 plus pages of music And of course many of the transcribers were inexperienced and they saw open score as a learning tool that They would do a transcription and then we would give them feedback on it But of course this is all basically what we're expecting and we even encouraged We wanted to get the The well-known works so we were hoping to get the longer works and and the way to Help those inexperienced transcribers as I published a set of Online exercises that they could complete before they started doing the transcriptions. So this would show them They would see Basically instructions of if you see this in the original score Here is how we would like you to notate that in Musescore. So it's like a step-by-step guide and We'd also help them by producing a template score. So we would add all the instruments To a Musescore file and set it up All ready to go so that they could just start adding notes from the beginning And the the rule for the template score is it has to contain all the instruments from all the movements And this is important when it comes to joining them up later And this is the first challenge is joining all of those scores together And we were relying on a Musescore 2 feature to do this called the albums feature, which was an experimental feature and There are various issues With this feature. So eventually it became necessary for me to write a separate program in Python that would take scores and Join them together outside of Musescore so that we could preserve as much information as possible. Otherwise some of the The transcribers work would be lost So then we would give the join score to So but the point of this is that even after the transcriptions complete There's still a significant amount of work that goes into tidying it up before it's actually ready to be published Another challenge is the variety of notation. So What we were hoping is that we could sort of gradually speed up over time. So we were starting out with that initial one Transcription being completed per month one of these symphonies, but It actually it was very difficult to speed up because we found that each new score would contain Sort of new notation quirks that we we hadn't encountered before the transcribers wouldn't have seen so we weren't sort of Learning over time and in the way that we hoped we would so I was learning but there'd be new transcribers and They wouldn't be gaining this experience in each new score posed new challenges. So the solution to this Was to actually in each template score. I would add Specific instructions for that score. So I would go through the The IM SLP score and look for things that I thought the transcribers might struggle with and then I would add instructions about those things to the score itself So this made work easy of transcribers It made work significantly easier after the transcription was complete But it created a huge burden in studying those scores before the transcription could even begin and in the extreme case and we have Wagner's devulcary here, which is the the opera that contains the riders vulcaries and So you can see the template score for this has a hundred and fourteen different instruments and that's because So what you're seeing here on the left is like a zoomed-in version and on the right you can see that line Indicating that's how much you're looking at so all of that down on the right-hand side of the screen. That's more instruments. So you're only seeing like the The wind section at the top and there's still the the brass below that and then the percussion and the strings and so on And this is due to the requirement to have every instrument that's in every movement And so there's instruments here like clarinets in a And clarinets in B flat and so on there's all the all the different keys and the score itself is over 700 pages long So this posed a significant challenge and So that because of the burden of creating these templates Meant that we needed to adopt a new approach So about this time we were approached by Mark Gotham from the University of Cambridge who had funding to run a project to digitize 19th century songs for piano and and a singer But we we didn't think we'd help him because we were struggling with our with our own project But I had another look at it and I thought well perhaps actually we can help here because these pieces that Mark wants to transcribe We're much shorter. There's two to three pages each So there's no need to split the score up and rejoin it later on So we just have one transcription group for all of the pieces rather than one group for each piece There's only two instruments in each score and they're always the same instruments so we just have one template that would work for every single one of these pieces and They're all based on a common theme so the notation was similar And what this meant is it was actually possible to get more people on board with the viewing With the reviewing whereas before it had to all be done by me because of all the specialist notation So this allowed us to complete 250 of those shorter transcriptions in four months Which is kind of what we're hoping for so with the main project and running our time is so speeding up we had we switched to doing transcriptions by individuals and So rather than a group would be one person would be the transcriber But this means that you the transcriptions run it in series rather than in parallel So one person working their way through the score so it takes significantly longer, but of course you can still do more than one at a time and So just recently 24th of December Muse score 3 was released and this has important features that enable us to Go faster in the future. So it has automatic layout, which means that The the task of tidying up a transcription that's been finished is much quicker And it also has a score comparison tool Which means that we can get printed out a list of if somebody edits a score we can see exactly what they changed and this protects the integrity of the scores and So we have the the planets here which Even though transcribers had done this on their own after the point at which it was given to us We had to keep sending it back to them to get it ready for publication in this to several months Even that was already finished But now we estimate that would take two to three weeks in musical 3 and I have a preview of the completed score here so you see the cover of visualization and We can Listen to the the playback on the score page on you's got up Well, you can't hear the play back. Okay. I'm very sorry to interrupt you But the time is up. We're actually over time already. So I would like to ask for a warm. Thank you for the talk