Bl labs sfu-dhi_lab-dhilab-2019-workshop

Editor's Notes

• #2: Morning everyone. <CLICK> I’m Mahendra Mahey, from the British Library in London, England, ‘Hello’. I am here to tell you my personal story about the experiences and lessons I’ve had working for my institution as well as with other Galleries, Libraries, Archives and Museums or ‘GLAMS’ at National, State, Public, University organisations and charitable and commercial organisations around the world. My story will focus particularly on how we have engaged with researchers, artists, educators and entrepreneurs from school children to adults who have used digitised and born digital cultural heritage collections and data to inspire them to create innovative, fun and inspiring projects. I would love it if my experiences can help other organisations build better ‘GLAM’ Labs, but I am also here to learn too from you. <CLICK> For the last 6 years, I have been running ‘British Library Labs’ a digital Laboratory to encourage anyone to experiment with our vast, incredible, sometimes totally unique and mind blowing digital collections and data. Our work has been generously funded over these years by the Andrew W. Mellon foundation and the British Library. We are in fact waiting for news any day now to see if the Library will be moving our work into its core business on a long term basis. <CLICK> During and after my presentation, please feel free to use twitter to amplify anything that resonates with you. My slides include much more information than I will have time to talk about, including links for you to delve deeper into the subject. On the right hand side in the footer there is a link to download all my slides, this will appear on all my slides. Please feel free to reuse them but it would great if you could attribute me and the Library when you do. My presentation will last about an hour. I will do my best to keep your attention. If there are any questions, something springs to mind, please make a note as. I will take questions at the end of my presentation, though I may ask you some quick ones along the way. So like all stories, lets start at the beginning and let me take you on my personal journey.
• #5: The picture you can see is inside the main building in London, it’s the King’s Library – King George the Third’s personal library, Mad King George! Sometimes known as the ‘stack’, I walk past this everyday and it gives me a sense of awe and reminds that the collections the British Library have are truly staggering and almost impossible to comprehend. <CLICK> We currently estimate them to exceed <click>180 million items, representing every age of written civilisation and every known language. Our archives now contain the earliest surviving printed book in the world, the Diamond Sutra, written in Chinese and dating from 868 AD….only around 8% of our collections are books and as you can see we have so much more, please note the numbers are only really guesses as to exactly what we do have. If you saw 5 items a day it would take you over 80,000 years to see the whole collection.
• #6: For me, this is what it is like trying to find a physical item at the Library. It feels like a huge hypermarket, or perhaps even a factory or warehouse. It stocks a random assortment of things and if you ask the assistants they can tell you about things that are simply not visible on the shelves in huge storage facilities.
• #7: Moving on to our digital collections which is where my work largely sits. What percentage/proportion of our physical collections are digitised? <CLICK> What surprises many people is that only an estimated 3% of our physical holding are digitised. This is because digitisation costs time and money and we have to achieve this through partnerships with commercial and other philanthropic organisations. <CLICK> Through one of the first BL Labs project, ‘Sample Generator’ we discovered that our digital collections are not truly representative of our physical collections. There will be all sorts of reasons why certain items get digitised and others do not. In reality, all our collections be them digital or physical have selection biases. Our collections are hundreds of years of decisions made by people as to which items to buy, keep and which ones to discard. <CLICK> In terms of licensing and using/reusing digital collections, a Lab like ours has further challenges. Out of our over 720 digital collections, only around 15% have an open license. The remainder are only available onsite at the moment. This is in part because many legacy digitisation projects didn’t always consider licensing when items were digitised. Trying to retrospectively establish rights and licensing on previously digitised collections costs time and money. <CLICK> As a National Library, we have been collecting born digital items for over many decades. We are the home of the UK Web Archive that periodically captures billions of UK websites to keep for posterity and research and we are the home of the Alan Turing institute centre for AI and data science where we are an active research partner. For example, Living with Machines is a five-year £9.2 million research project that will take a fresh look at the well-known history of the Industrial Revolution using data-driven approaches. <CLICK> A new digitisation programme Heritage Made Digital is trying to learn from past digitisation projects, especially on digitising collections based on research demand.
• #8: It can often feel like this…It’s much smaller, we have some free stuff, some can only be consumed on site, some you need to buy. If you speak to shop keeper, they may be able to get you to see what’s under the counter, because they couldn’t display it. You might be able to get special permission to get a look in the warehouse at the back of the shop which has even more goodies there. If you are looking for vegetables you have come to the wrong shop!
• #10: How do you find our open cultural heritage collections? On way is to use our collection guides which offer subject pathways into our collections, each guide will have a section non what’s available digitally if at all.
• #12: What’s important to understand is that if you really want to work with our digital collections, it sometimes pays to learn the ‘back story’ of how the collection came about, this was a really early and important lesson I learned. Knowing it, can have a significant impact on what you might want to do with it. On the screen you can see many factors. I simply haven’t got time to go into them all, but perhaps the most important one is the last one. Is there a human being around in the organisation who can tell you about the collection, as communicating with them may be the quickest way to learn more about the digital collection you want to work with. Often, they will have access to important information that isn’t written down.
• #14: As a Labs manager, I faced a significant challenge of how we would enable those who want to access to our 85% of digital collections that are only available onsite at the moment. If we look at this, we can see onsite at the Library may mean that the digital materials are only available in the reading room on a specific PC, or that the materials are still on their original storage media and may need obsolete equipment to access them or they still are as yet to be transferred onto a more modern system. <CLICK> Some digital materials are only available through payment <CLICK> Only a small fraction of digital materials are in the shiny happy carefree open web (more about how to access these later) <CLICK> What we developed to tackle this situation was to develop a ‘Residency Model’ initially through an annual competition that we ran and now this has evolved in application process where researcher’s can apply to carry out digital research onsite at the British Library. These researchers in residence have special access to digital collections that our regular readers do not have. Access is strictly controlled depending what they would like to access and what they want to do with the materials.
• #15: At the Library, interpretations vary of what ‘onsite means’, does it mean only on a PC in a specific room, on all PCs in all reading rooms on all sites, <CLICK> In order for Labs researchers to be resident, they need to be security cleared to gain access to specific collections . Once done they can get access to a hot desk in a staff area, and we are trailing something similar in the reading room in a secure space, sometimes they can get access to their desktop remotely too via Citrix and Virtual Machines. <CLICK> Sometimes further permissions are required even after security clearance depending on what they want to get access to and what they want to do with it <CLICK> Digital material can be on various media formats (not always online) and they need to be mounted on obsolete devices or transferred onto more modern equipment <CLICK> We are getting better at providing access to our digital collections but we still have a lot of work to do.
• #16: 24 seconds (72 words) Let’s look a little further at the types of interactions we have with our researchers. We have summarised these phases as ‘Exploration’ where people often ‘rethink’ their ideas of what they want to do with the data, ‘Query-Focused’ where they often have to iterate to come up with a realistic proposal of what they want to do and a ‘Wrap-up’ phase to end their project with us, if it is relevant.
• #48: 970 files from a selection of 19th century newspaper titles from the BL corpus for us to correct using the overProof post-OCR correction software The best way to measure the improvement made by the correction process is to compare the OCR'ed text and the automatically corrected text with a perfect correction made by a human (known as the "ground truth"). Hannah-Rose's 5 small human-corrected samples are show as green dots. These are not only smaller than the other files, but their raw error rate is much lower at 13.3%. OverProof was measured as reducing this to 5.4%, a removal of almost 60% of errors. The red dotted-line indicates the correction "break-even" point: the further under the line, the better the quality of the document after correction. In the graph below, the grey line shows distribution of files across error rates before correction and the green line after correction.
• #63: Posts small illustrations taken almost at random from the digitised book corpus to a Tumblr blog. This experiment with undirected engagement was a by-product of work to uncover the hidden wealth of illustrations within the digitised pages.
• #82: 50 seconds Here is the anatomy of a Flickr record, importantly we have created links to many of the Library’s services <click>some of this lovely traffic is going back to the Library and hopefully generating more interest in our services, from downloading a pdf of the book to purchasing a high res scan of the image. <click>Tags are added from the original book record, including the approximate page number the image came from<click>users of Flickr can add their own tags, and I have mentioned they have already started doing it.