 And first up on this panel is Joe Knuckles, who has various affiliations. One of them is the University of Edinburgh, the University of Glasgow, the National Library of Scotland, and he is also a satellite member, you could say, of the co-operative. So yeah, very well-established in the academia, at least well-known there. Good. And the Talkers of Titans, how to best facilitate our work, study of transcribers of free processing requests, people are putting the free credits that are available through our scholarship programmes to good use in practice. Then the floor is yours. Thank you. Thank you. Thank you for your introduction. Thanks everyone for coming. And it's good to be finally physically present in Innsbruck, three years into my current PhD, of which this is hopefully going to end up as a thesis chapter in some sort of way, shape or form. We had yesterday the privilege to hear six presentations from those who've benefited from the free processing scheme or the transcribers scholarship, which is how I'll refer to it during this presentation. This presentation attempts to take a broader lens to kind of show the extent to which the free processing scheme is being used, taking more of a quantitative approach. It also attempts to address the broader question of to what extent free processing initiatives from digital humanities platforms are supporting early career researchers, students, and those with limited funding. As we're all aware, especially those who are within university institutions, it's clear that financial support is necessary to increase the diversity, but the quality of DH research as well. So as such, this paper, this presentation helps clarify who is making use of the free processing scheme, whether the scheme currently is helping mediate access issues in a sector which is increasingly becoming more unequal, sadly. So our general trajectory for the talk, hopefully the slide. So our general trajectory, we're going to focus on three main things, depending on the time and how much I elaborate and talk. So we'll provide a bit of background to the transcribers scholarship and the free processing scheme, which I'm sure some of you are familiar with, especially those who presented yesterday, will then present the findings from quite a large content analysis of over 150 request forms, which was sent via email, and also kind of blend that with a supplement document, which was drawn up by the main contributors who help with the free processing scheme over at READ. I'm also going to present some preliminary findings from a current survey, which is really bad practice because the data collection hasn't ended, but I'm going to give you a kind of insight into that ahead of time, and just some roundup thoughts as well. So DH as a sector is now recognising the historical barriers present in accessing digitised material, digital tools and digital infrastructures. This inequality goes well beyond financial barriers. It raises issues of language, hardware requirements and skill barriers, both as isolated factors and combined as this toxic brew. As a result, there's been a healthy weighting historically in digital humanities towards the global north, and that can often be seen in the sector. And to name-drop a really, really good researcher who's working in this area is Rupika Rizam, who focuses on kind of coalescing a global south approach to DH to correct that balance. So these issues present the need for a data ethics approach, compiling counter data in order to name impression when seen. As a response, those in DH should always seek to centre social justice minded approaches, particularly when building tools, designing new workflows and considering means of access, challenging these previous inequalities and resisting the reinscription of exclusionary behaviours. Though mindful of the fact that inequality can come in various shapes and forms in DH, this presentation is going to look purely at funding, answering whether the Transcriber's Free Processing Scheme can be part of this social justice approach, making DH a sector more equitable. The paper will also seek to address the following questions. What are the demographics of those using the Transcriber's scholarship? Again, we saw six great presentations yesterday, but what work is the scheme enabling? Are these schemes being utilised by the intended groups? So early career researchers, students and those with limited funding and how can the scheme be extended to enhance this equitable access to DH platforms? So as most of you are aware, just to give a bit of context, on October the 19th, 2020, Transcriber's transition from a free to use tool to a pay for model with HDR training corresponding to a required amount of paid for credits. This formed a sort of freemium structure, although that model doesn't quite fit where Transcribers' base functions, layout analysis, the text editor to perform your ground truth transcriptions were free to use with additional features, HDR training coming at a cost. Although the success of freemium models, especially in the business community, is unclear. Such strategies are well used by software companies and are seen as a way of moving beyond one of customer interactions towards long term engagement. Newly registered accounts of Transcribers, I'm sure you're familiar with this as well, receive 500 free credits. And that's, you know, roughly, approximately in my own work is kind of enough to run a model over 400 to 500 handwritten pages. It obviously varies if those pages are printed text as well. So Deng in 2022 stressed that the free version may allow customers to sample the product before making a purchase decision and subsequently increase demand for the paid version. But it might also cannibalize demand of the paid version. Transcribers, in this case, is well positioned because users can train a model, you know, experiment within the tool and work out actually if the tool is for them and important to their work. However, in the case of individual researchers working on small collections, these free credits could be enough to train a model initially for, you know, simple searching. Therefore, whether users purchase credits from Reed is very much dependent on the size of their collections. So actually, what's behind me, those applying to the Transcribers scholarship program must fill out an online form. They have to indicate the amount of credits they expect to need. Their home institution, although this can be left blank, has no bearing on the process and a short description of their work, which is where it gets really interesting. If accepted, credits are then allocated in an email notification ascent. Credits can be checked and monitored through the online portal of Transcribers Light and applicants can request up to 3,000 credits. We'll get more into that as well. So the broader study formed a content analysis of emails generated by the scholarship system, which looked like this. This is what I received every time someone made an application to the scholarship program. So it's almost kind of 20 emails a day on top of everything else. So I got quite used to checking them. And the emails that we've kind of factored into our analysis were sent between the 7th of November, 2020 and the 16th of March, 2022. So it broadly encompasses a year and a half worth of credits. And this allowed for the exploration of individual cognitive processes of applicants related to the characteristics of their requests. So students, those conducting workshops and those with limited funding were quite easy to identify. Early career researchers were harder to pinpoint because a lot of people didn't elaborate on their position within their given institution. In total, 161 requests were collated and interrogated. These were then aggregated, anonymized in accordance with the University of Edinburgh's GDPR guidelines. And alongside that, a set of six semi-structured questions were sent to Reed staff, which they kindly reply to in a very well-organized one document that was quite easy to read and follow. So those responses have acted as a supplement to this study. So to actually get to the findings of the second bit of our presentation, the free processing requests submitted to the Trans Groupers team came from a total of 99 institutions. It's not surprising that most of these institutions were universities and research institutes, 91 out of the 99 roughly. But it gets quite interesting when you dig deeper into the additional eight institutions. Two free processing requests came from high schools, high school teachers, one from Turlock in California and one from Vienna. Elsewhere, national and state libraries were involved, publishers, and the Royal Botanical Gardens at Kew who were looking to apply Trans Groupers to some of their field notebooks. It appears then that Reed is reaching their main demographics, content-holding institutions, university researchers, with the scheme only being used beyond these groups to a minimal degree. The publicity and targeting of the scheme appear to coincide with the aims listed by Reed's staff of facilitating research and contributing to the spread of methodologies and the history in the era of historical documents. So looking at the dataset geographically, applicants came from a total of 31 countries, Germany, the Netherlands, the UK and Switzerland are the best represented. But compared to previous quantitative studies of the Trans Groupers user base that we've completed, mostly of materials found on Google Scholar where Trans Groupers is mentioned in scholarly research. This appears more diverse than those, that formal published research. So more work geographically seems to be happening in the remit of informal experimentation, research tutorials. They seem to be distributed a little bit more widely. So Reed's focus on enabling common transnational activities is clearly still a work in progress, although more diversity is seen in these requests done in those cataloged research papers. Despite a lot of these nations recognizing English as an official language, we had free processing requests from Australia, Brazil, Canada, Ireland, Egypt, Estonia, Greece, Hong Kong, New Zealand, Norway, Portugal and Slovakia. All of those were represented in these requests that weren't represented in scholarship that is easily found online through Google Scholar. So Trans Groupers appears to be at a center of transnational practice when looking beyond this scholarly research at grey and material like credit requests, moving away from purely a Western European view of digital humanities and this global north perspective that I mentioned as well. That said, the caveat is amongst the 161 requests, 140 came from European institutions, so 87%. So we've got a long way to go before Trans Groupers is used worldwide. The range of applicants domains was also broad, reaching across different trading zones between the humanities and sciences and the variety of research interests is hard to summarize in this presentation, but we had people working on the history of the 17th century Dutch Republic investigating poor relief in Zurich in the 19th, 20th centuries, someone working on the 18th century burning of Kingston, Jamaica, the 19th century Irish potato famine, histories of Jewish families, and that's just to name a few. Others had kind of more space to offer more insight into how greater access to transcribes was solving issues in their work. A lot of people struggling with Roman cursive and current drift who have kind of seen transcribuses as a major help in their research to make, as Reid says, the past readable. So in taking into account language materials, the diversity is even clearer. Free processing requests range from work on ancient Aramaic, Arabic, Peruvian administrative documents, 16th century Slovak evangelical sermons, parliamentary speeches in Scots, the Sengali's written press check, Spanish literature, a huge amount of stuff. I'm just going to check the time so I'm not running over. Am I doing okay? Thank you. I didn't realise there's a card. That's quite snazzy. We're okay, I think. It can be identified that the transcriber scholarship is being used by the intended groups. 70% were students ranging from the undergraduate level to PhD candidates. Early career research has made up about 3%. So it's actually quite a small number. And those performing tutorials made up about 15%. Although not prevalent, a few credit exemption requests came from quite established lecturers in the field. Whether that's because maybe they misinterpreted what the free processing scheme was attempting to achieve, or whether it's an indication of the current state of academia that even established lecturers need help from the transcriber scholarship. But one read he, the applicant, is the author of more than 100 scientific publications in newspapers and conferences. And one has an edited book. His research interests include pattern recognition, machine learning, deep learning, and image analysis. So yeah, quite interesting that even someone of that high esteem is still using the transcribers scholarship. Just being conscious of time, I'm going to throw some more stats out. Broadly, if all the applicants gained their required amount that they enlisted actually in their applications. Although many asked for more than the 3,000 limits for credits, transcribers and read have deployed and offered a total of 666,882 credits. And using the calculator on the website that corresponds to about 128,000 euros. So it's a sizable chunk of money that's gone out to ensure that people have access to such a wonderful tool. One of the major things that came out from the review is that applicants sometimes struggled with figuring out the amount of requests that they need. So a lot of people are uncertain. I don't know if 10K credits will be enough about dot, dot, dot maximum. I'm assuming I won't need more. I really don't know how many credits to ask for as a reasonable number to distribute while giving workshops. Advice would be welcome. Read staff also highlighted that sometimes working out credits could be difficult. Pressing perhaps the need for streamlining that aspect of the service. Although the current structure enables longer discussions between read staff and researchers, usually via email and any streamlined structure could potentially jeopardize that sustained contact with our user base. So just to plug the next step quickly, again being conscious of time to kind of add body to this quantitative study, we want to hear from you if you've benefited from the Transcriber Scholarship and if you've managed to evade me so far because we have sent multiple emails out. So hopefully the slides will be made accessible and you can find the enclosed link. The survey began in August 1st. It's due to end. Hopefully don't leave it too late on January 1st, 2023. I'm not sure I've got much time to actually tell you any of the preliminary findings. Two, three more minutes. Thanks, Andy. I don't want to deprive everyone else. So a lot of people have said that without the scheme, you know, their work would otherwise not be possible. In some cases, the sheer scale of their work was cited as the main obstacle prior to gaining financial support. So free credits helped me transcribe and scan through larger quantities of material than I could have physically. Another person said, without receiving free credits, I would have had to minimize the text corpus to complete my research, which would minimize the outcome and its reliability. Even those who had enough time to complete large amounts of work or were even willing to pay themselves thoughts similar with granted credits, which would make processing easier and faster. Someone said yes, but it would have taken longer to process and I would have had to overhaul my methodology. A few people, again, as perhaps in the document of current academic funding, hinted at more structural problems at their institutions. Someone said there is no funding available for postgraduate talk dissertations at my institution, so that money would have had to come from my own pockets. Another cited that the prospect of gaining any sort of financial support for student-led work was an impossibility at their institution, which made for quite sad reading. The vast majority of respondents, 78.95% were happy with the scheme, offering no suggestions. Some improvements were mentioned, mostly directing more training materials towards those who are unknown to have received funding from the scheme just to reduce the learning curve when people first use transcriptus. As a follow-up, just as a brief visualisation, this goes back to the free processing requests and not the survey, but here are the most frequent words used across all 161 requests. The size corresponds to how often the term was used. Of course, it's no surprise that the biggest term is transcriptus, which is mentioned 148 times, but elsewhere you have some quite interesting findings. I realise I've still got iron, which I probably should have removed as a stop word. But student, university, PhD studies, which all have a clear association with students applying for free processing, again show that readers is probably meeting its current user base, its intended user base for the free processing scheme. So just to finish off, to ensure that DH remains, well, is as equitable as possible and that the re-co-op remains sustainable, having a solid dialogue with our users is essential, whether it's free, free processing, free surveys. So as such, handwritten text recognition requires new approaches both to historical material, to public engagement in order to best support our intended communities and support the ways they're applying transcriptus to cultural heritage materials. So that's far, far, yeah, certainly enough from me that I'd welcome any questions. And thank you for listening to a lot of stats, but hopefully it wasn't too dry. Thanks a lot, Joe, for these insights. And yeah, this is really a very important aspect of the cooperative to be able to support early career researchers and also people offering courses and trying to, yeah, sort of educate the next generation of researchers what's possible these days with the help of AI. So are there any questions related to the talk? Just a question about, did you take into account the big projects like ours that by bulk soft credits that benefit directly the student researchers? I haven't, that's a really good, really good point to make. This is purely a study of requests that were made by the students themselves. Often you get supervisors who are quite supportive of obviously what their students are doing. That's certainly another step we could take, definitely. And then the stats are even more impressive for the sheer amount of credit set of being. And also we asked in the summer schools, we ask transcribes to like 40,000 credits to run this school in the summer school or the course also, you know, that we are giving. So this gives another level. Yeah, we have done work, a survey went out last year that was the second major survey of the trans group as user community after Melissa Terrace's one back in 2019 that looked at institutions instead of individual researchers. So this was kind of an attempt to get to the nitty gritty more individualistic studies. So hopefully between those studies we somehow captured the actual makeup of the cooperative. But it's a real, really good suggestion. And the nice thing here from a process kind of view is that for the scholarships, we have a very good database, you could say. And so there you really know how many scholarships have we granted, how many credits have gone out with projects where we, yeah, provided some free credits in order to facilitate, for example, summer school or something like that. So I don't think we have a central record of those. So those would be harder to find out about. So that's just a one, a pragmatic aspect of this study. I think you should say as well, the biggest amount of credits that was applied for, but obviously read. We are quite close to the end. Obviously read couldn't give their required amount, but we did have a few applications that came from very big distribution-wide projects. So someone asked for 30,000 credits. I think was the top end. And of course, a big dialogue happened. So we've still got an indication of some of those projects happening. Yeah. And we've also supported researchers with larger projects in emerging economies, at least for part of the project. So we've done that too, which did not happen strictly. Yeah. Under the flag of the scholarship program. But so, yeah, this is basically just the numbers that you are able to get in a clean way. So yeah, maybe we should keep better track of the others as well, because they are impressive too. So, yeah. So that's more our fault and no fault of yours, of course, for not keeping a good track of those numbers. Any other questions? I think we've got time for maybe one more. Otherwise, we can move on to the next talk. So, sorry. Thanks a lot, Joe. You got a transcript with mag two. Feel free to fill it with tea, coffee, or IPA. I hear. Thank you.