 Oh, I need the USB-C to HDMI. Oh, you don't have one here. The other room? Maybe the other room. I don't have one here. He will grab one in the other room? Okay. Okay. Otherwise, I'll be able to put my... I guess you can just put my slides up and give me a quick... Yeah, I can... Okay. Shoot. So what I do is I need a USB-C to HDMI. USB-C to HDMI. Let me go check from the other room because I had one the other day. Thank you. Thank you. I'm sorry. And are you going to need like a clicker? I can just use this. Okay, good. So let me just open the email. And I have the rights to... Is there a right to edit it? Is there a right? It can be all rights. And is it okay to just mirror the screens? Yes. Mirror screens. Okay. Can you see my screen? There we go. Okay, so let me go full screen here. All right. I'll go through this. Excellent. Okay. Wonderful. Thanks a lot. Okay. Excellent. So folks can come on forward if they like. You can see a little bit better. Some of the text might be a little bit small. Perfect. So I'm James Heilman. I'm with WikiProject Med Foundation. And for this here talk here, I'm going to be discussing the relaunch of our translation effort. So thank you all for joining. So a little bit of history around what we'd be doing. The healthcare translation work was something we started back in 2011. So it's been going on now for a dozen years. We initially began with partners at Translatorsville Borders, among others. And the first process was a little bit cumbersome, but we did manage to create 19,000 translations into other languages, representing more than 22 million words of text on various languages of Wikipedia. We saw work happen in more than 150 languages, and there was more than 3,250 translators who took part in these efforts. With respect to the translations that we created on these key healthcare topics, they've received about 3.7 billion page views since 2015. And this represents, you know, translation is a very common method that we get healthcare information in other languages of Wikipedia. Translation represents about 40 to 50% of our healthcare content that exists in non-English languages. So the initial starting content was 1,300 articles. And what we're doing is we're improving the leads of the English Wikipedia articles, making them a good summary of the topic in question, and then putting those out for translation. So just those three to four paragraphs of content, we're encouraging people to translate into other languages. The English articles themselves received about 4.5 billion page views, those 1,300 since 2015. So over time, we generated more and more content. There's more and more content for us just to keep an eye on and try to keep up-to-date on English Wikipedia. And, you know, we were doing much of the moving manually. You know, we're a cloud-bringless trend, there's no borders. We were creating Word documents from Wikipedia articles. We're emailing them to translators. They were giving us Word documents back. Wikitext will get broken in the process. And then, you know, it was a lot of work for us as volunteers and coordinators to get this material back on the language in question. So it was just becoming too burdensome. And, you know, over the years, English Wikipedia has become more and more unhappy that it's having references in the leads. English Wikipedia says that, you know, the references should be in the body of the article and the leads should not contain references. And that makes it difficult for translation when all you want to do is translate these short summaries into other languages. English Wikipedia is also unhappy with us using easier-to-understand language. You know, the language level we're trying to write at is about a grade 12 reading level. And English Wikipedia is looking more for, you know, sort of a grade 18 reading level for medical content. And then finally, we lost some of our key contacts at Planetaries of Borders. And then, you know, monitoring English Wikipedia content was simply taking up almost all of my time. But what did we do? You know, we took a step back and we changed a bunch of the workflow and we relaunched our translation process here in 2021. One of the key things that we started our own website, mdwiki.org, where we could write these short summaries of key healthcare topics and then take these key healthcare topics and translate them into other languages of Wikipedia. So the final location for the content is still Wikipedia, but the starting location for the content is our own website, which gives us a greater ability to include references in the lead and to use easier-to-understand language that is easier to translate into other languages. Additionally, we've built a dashboard that has automated much of the process of managing the Wikitext and our dashboard interacts with content translation, the tool built by the Wikimedia Foundation to move content between one language of Wikipedia and another. Additionally, this has given the opportunity for us to fix a number of the issues that content translation struggles with. One of the issues content translation struggles with is that it drops a bunch of references and they weren't quite sure how to fix that. So one of the workarounds we've built, this new process is allowed at the build, is we expand the references and then we load that content with all the metadata expanded and then we shorten the references after it reaches the target language and that has allowed us to keep the references going missing if you've noticed that problem with content translation before. Another benefit of having our own tool, our own dashboard that has automated the ability for us to not only collect the impact of our work but also it has allowed us to follow along and find translators who might be running into issues in the target languages they're working in and provide them guidance around some of the technical difficulties. Additionally, it's allowed us to bring on a number of new partners including working with the WHO around hearing related issues. We're working with NIOSH around some workplace safety issues and we have a new volunteer translator group called prosed.com. So as I mentioned, we launched about two years ago since we've generated about 1,500 more articles ready for translation in other languages. We've had 53 translators join us. We've created almost 1,600 translations which represents a quarter of a million words. We've seen work in about 30 to 50 languages and this content has received about 1.5 million page views. This is an example of the type of material that we have put up for translation. The text is a little bit small but what you see is here's a four paragraph summary. We're only recommending people translate. Only this amount of content gets loaded into the content translation tool so it helps guide our translators and what amount of content we're expecting them to work on. One of the nice things about this is this is 300 to 400 words of text. This is something manageable. It's not something like taking on a 10,000 word maths of English Wikipedia article. When we started our translation efforts we got feedback from some of our smaller communities and we initially were translating entire articles and they simply found that that was too much content for these small communities to manage on an ongoing basis. Additionally, further research from the Wiki Media Foundation has found that only a small portion of people read beyond the lead of the English Wikipedia article. 60% of people never make it beyond the lead of the English Wikipedia article which tells us that if you want to have a big impact on our readers make sure that the leads of these articles are understandable and are correct. Let's take a look at how the tools work. Where's the plus key here to make it bigger? Found it. Here's a look at the landing page of our translation task force. We tell a little bit about the background and then we have step-by-step instructions on how to translate your first article. The starting point is our translation dashboard. This lives on Toolforge and let me see if I can make that bigger here as well. There we go. Basically, what we see here is we have campaigns and the campaigns are based on categories that exist within MD Wiki. We have our main campaign. We have a campaign related to COVID. We have a campaign related to hearing which we're working on with NIOSH. We have a campaign with respect to essential medicines that we're working on with respect to the World Health Organization. What you can do is you can pick one of these campaigns. Let's say essential medicines and then you can put in your two-letter language code. If we pick Japanese, for example. Oh, I need to log in. Sure, you made it as well. Yeah, exactly. So we have I-Off on MD Wiki and that basically allows you to use your Wikipedia username and password along into this system and it keeps all of your edits associated with you as you move through the content translation process. Okay, once more. You might need to hit log in with Wikipedia. Yeah, I did. Oh, maybe just hit log in. I did both. Oh, okay. Want to try your credentials? I think I have two doctor authentication online. I will try one more, but I think I'm biting this right. Maybe hit the top one. Okay. So what the tool will do then do is it will generate a list of what your articles are missing in your language that are ready for translation into your language. Okay. Sorry. No worries. And then from that list, you know, you'll, oh, and then from that list, you'll be able to then pick articles and it pretty easily walks you through the process. The other feature that we've built is we've built ourselves a dashboard and this gives us a breakdown of how people are doing with respective translation. You know, we see Odea is leading the, leading the board with the director the number of translations done. This retired orthopedic surgeon has translated more than 1000 articles since the relaunch of our efforts. We have the number of languages we're working in and then we have a graph of number of translations over time since being launched back in 2021. And then we have some additional tools here that sort of, you know, allows us to pick up quickly on new translation that are occurring. And so what we see here is, I can make it a little bit larger. These are the most recent translations that have occurred via this tool. We see this JSDS has been translating copper IUD into Spanish and you can easily take a look at the translation that result. So to see whether or not this user has run into any issues from the Spanish community with respect to the resulting translation. And then we additionally have built tools that help fix a bunch of common issues. We're still working on improving those tools. So for example, when you translate from English into Spanish the dates of the references, January's not translated into Jean Vieille or the Spanish equivalent. So we're building tools to help automate the bits of text that don't translate automatically with the content translation tool. And then we also have the ability to view what translations are in process. And this, you know, some people start translations. They run into barriers. And you're not quite sure why. And this gives us a list of people who have begun translations but maybe haven't finished them. And this allows us to reach out to those folks and see if they need further help. So we've built these tools to help people run translation campaigns. Currently we have campaigns primarily related to medicine and healthcare topics. But there would be definitely the potential to run campaigns and other topics using this sort of software, this tool here as well. Oh, hey. So, yeah, so as I mentioned, you know, we also did a big effort to look at what sort of impact our prior translations had. One of the nice abilities allowed us to figure out who our prior translators were. And it allowed us to figure out what percentage of all of Wikipedia's medical content came from these translation efforts. And it was more than 40%. And then we recently reached out to all the people who previously did translation as part of the healthcare efforts and encouraged them to rejoin us for this new process. Finally, I'd like to put out a big thanks to Mr. Ibrahim. He is our part-time programmer. So we have Wikiproject Men's Foundation. We have one part-time staff member. That's this gentleman here from Ibrahim. From Yemen, I should say, Mr. Ibrahim, who's done much of our programming for our tools. And then, of course, a big shout-out to Dr. Chandra Root, who has been our star translator and has, you know, written. Odea Wikipedia has some of the best medical content within any language of Wikipedia. And that's because there's one single gentleman. And he's basically brought a medical encyclopedia single-handedly to his language, which is an Odea language not known to many people, but it's spoken by 40 million individuals in India. So these sorts of heroes continue inspiring me. Okay, I'd like to open up for questions. Let's bring you a microphone and just so you can repeat that question So I think that the main concern for me regarding the project that all of the project depending on the translation tool. So any issue in the translation tool will affect the project in a negative way. In my part, I have a lot of concern regarding the translation tool in the Arabic language. So I still use the manual editing way to translation or try the medical articles in Arabic language. So that is my comment only about. So there's two parts to the translation tool. One part is just managing all the reference templates, the info box templates and helping to convert those from one language to the other. And then the other part of the content translation tool is the machine translation. And it's the machine translation that concerns it. The machine translation is not very good. And that's why we, you know, human translators are 100% key, especially for medical content and why our work is so important. You know, you can go to the internet and you can use Google's machine translation to translate any English page into Arabic. But the resulting content that Google gives you is poor. You need real human Arabic speakers to come and, you know, take high quality English content or in other languages to generate high quality Arabic content. And, yeah, so, you know, in Japanese, for example, they have concerns about machine translation too and they have actually turned off machine translation for all of the Japanese languages. Now that sometimes it's kind of easy for the ordinary people to see the Arabic translation presented in the medical class. I'll tell you the last example today that Google is considering as only a Russian translation presented in Arabic cannot be functioned to do the cardiac as a heart and the cardiac as the organ of the stomach. They translate all of it as a heart. So when you read the translation, all of the content related to the cardiac related to the stomach are translated as the heart, the heart in Arabic. So when you read the text, it's a very poor person. And you need a specialized one you can say or one related to the main content that you use to kind of differentiate between... Yeah, and part of the solution to that is, you know, we're writing summaries in English and we're trying to use easier understandings. So, you know, we wouldn't use terms like the gastric cardiac or the English version. And if you avoid terms like gastric cardiac, then if you say the upper part of the stomach, then the machine translation will be able to handle the upper part of the stomach whether they won't be able to handle the gastric cardiac. And that emphasis needs to be on the English folks who are writing the initial content to make sure that we're writing content that is easier for not only non-expert to translate but also for the machines to work on. And yet, you know, if you can provide a feedback on specific issues that are occurring on the English easier simplified version, that would be useful for us. One moment. One's out. Oh, okay. Hello. I can see, like, this MD Wiki is trying to be independent from Wikipedia, is it? The MD Wiki is trying to be independent from the Wikipedia. Is that right? So, you know, we're, of course, a movement. We're the Wikimedia movement. Wikiproject Med Foundation is a somatic organization within the Wikimedia movement. MD Wiki is a Wiki run by the Wikiproject Med Foundation. It's not run by the Wikimedia Foundation. You know, that sort of distributes some of the, you know, technical infrastructure across our movement, which I think is, you know, a good idea for us, you know, as a whole for stability for the future. You know, this is independent from English Wikipedia. This does allow us to create more specialized versions of articles. But the resulting translations go into Wikipedia. So, it's separate but still related. So, how do you ensure this contents in the MD Wiki reach out to the readers? Yes, I think if you Google on the topics of any health or medical things, the priority will be from the Wikipedia. How do you ensure these are the contents from the MD Wiki to get to the people? And, you know, that's via translation, right? So, a content from MD Wiki is translated, as we said, there's 1,600 translations into other languages. So, you know, we're seeing more high quality content generated in languages other than English, which I think is key. You know, as an English speaker, more than 60% of the internet is written in a language that understands. You know, that's not the case for so many non-English speakers. So, I believe we need to put more effort into languages other than English. And this is part of what, you know, the family effort is fun to do. All right, thank you. MD Wiki to bring visitors. It's only for to help translators. So, it's not to bring visitors in MD Wiki instead of English Wikipedia. Yes? So, is that the aim? Yeah. It's not the aim to make a separate project MD Wiki for the medical articles and to bring the visitors to it. I think it's not the aim. The aim to help the translator and to avoid the guidelines and policies of the English Wikipedia related to the medical articles only. Thank you. Oh, yeah, sure. Or I can use this one. So, you know, with respect to is this competing against English Wikipedia? You know, our offline app in English, for example, is now based off of MD Wiki. And the reason why, you know, we know that 80% of people who are using our offline app are from low and middle-income countries. Many of them are healthcare providers. And many of them are using it in a location where there's no internet. And some of the things that I've been requesting is, for example, medication doses. Medication doses aren't allowed on English Wikipedia. You know, us having MD Wiki allows us to then provide medication doses to our offline users. So, this allows us to do a number of things. You know, it allows us to improve our offline offerings to healthcare providers around the world. You know, it's not directly competing with Wikipedia because, you know, we're not well indexed by Google. Google is not sending us much traffic. And I don't think that's going to change. Yes, the only thing that I want to add is that for the translation tool, the main improvement than any other machine translation as I tried it is that it used the inter-wikilings. So, for example, when we go to a term that is in Arabic Wikipedia, if the Arabic name in Wikipedia is correct, then it will use the correct name. But other machine translation, they don't go to a reference. They don't go to Arabic Wikipedia to see what's the name of the article, for example. They use the general dictionaries which do the problem that my colleague raised it around, like, for example, cardio. And there's other examples. For example, intelligence as a smartness intelligence as, you know, the policeman or the security man. So this kind of term that machine translation usually does not differentiate between. But for our tool at the MD Wiki, I tried it myself. There's still some mistakes. It's not human. But it is good quality compared to any other, like, Google Translate. Because it used, it's referred to Wikipedia articles. Other thing for the visitors, the outreach as a question raised by our friend there is that I want to mention that MD Wiki is connected by inter-wiki to other wikis. To English Wikipedia, for example. So this is another way to reach to MD Wiki. But we are not in a state to compete English Wikipedia, but rather we are trying to provide simplified content, more accurate content in our point of view. For example, as Dr. James said, there is, we add the prices. The prices, for example, are important for certain parts of the world, which is we cannot add to English Wikipedia. That's why we have this version. And also for internet and a box, if you attend the session yesterday that Dr. James talked about internet and a box, the offline version of MD Wiki is now available in more extended way, I could say, than English Wikipedia articles. And also we have other educational resources, like videos from Khan Academy, from Osmosis, at a certain point we add Osmosis videos. So this is what I think is not a competitive thing, but we are trying to add something different. That's it. Hi, I'm a doctor too. So, but however, I quit translating to Arabic, medical content to Arabic from English Wikipedia a long time ago because of the complicated content. I believe now science is highly complicated that I get to the point that I write the same word in English just with Arabic letters, so I always say what's the point? I'm not explaining anything, no one is having any benefit from that. So, having a simplified language is better in my opinion, but surprisingly this is not getting any traffic so what's the point? Why would someone contribute in this project if it will not have much audience? And my other point is that I'm not a programmer so I try sometimes. I hope I will get there someday. But is there a way to prevent publishing something before it gets a certain percent of editing? So, we make sure that it's not just machine translation. No, there is some manual and some person that has been done so, yeah that's better and can be published. So with respect to readership the translations that have occurred, we've seen that they've so far received 1.5 million page views in non-English languages and 1.5 million page views, that's not horrible that is some impact and of course those numbers are going to grow over time. These articles have existed for less than two years and yet people are still actively reading them in these target languages. I have a question myself. I'm wondering as you presented this like the main version of the article is in English and everything gets translated from English. Is it considered a different way of doing it, like for example somebody writes a very good medical article in Spanish and have it translated to other languages? So, we were previously translating into the language quicha and our only translators only spoke quicha and Spanish. So we ended up doing a two-step translation process. We ended up translating from English into Spanish and then from Spanish into quicha. One of the big efforts of this project was to make sure that the starting content is of very high quality and because I'm an English speaking physician I can only really verify the content in English and most of the few other physicians who are working on developing the content, the starting content are also working in English. That's more or less why we start in English and when it comes to medicine the lingua franca of medicine is very much English. Globally I did some time in Brazil as a medical student and the medical students were studying English even though Brazilian was their first language same in Switzerland even though English is not one of the official languages of Switzerland the medical students in Lausanne Geneva or Switzerland were still studying in English and that's the same in much of India and it's interesting like one of the push backs we got from the Swedish community is they wanted Swedish sources to be used for the Swedish medical content but the thing is their own researchers researchers in Sweden are publishing in English so even the Swedish experts aren't writing in their own language and so translation is just a fact of life. This is a question I guess about synchronization I can totally see the reasons for forking essentially MD Wiki off the initial starting point of medical content from English Wikipedia and then it gets integrated into other languages multiple seeding articles being created all the time and the question is then as material updates over time even some topics update very slowly but some topics update quicker and even for slowly updating topics in five years time in ten years time this will be relevant how are there plans for then updating and synchronizing the language the other language materials that have been generated from English content and also synchronization between updates to English Wikipedia into MD Wiki updates in MD Wiki back into English Wikipedia is this all entirely manual is there potential for some AI assistance even at least for spotting what could be synchronized even if it's then having to be manually checked yeah you know so right now we're only listing articles for translation when there's zero content that exists in the target language you know if there was four sentences of unreferenced content in you know a target language and you were to take three paragraphs of well referenced translated content from English and put it in that language that is sort of an insult to the community in question and you know you're stepping on their toes and really you know that sort of level of work can only be done by the core members of that community and can't be done by us or external translators that have joined us from you know our various collaborators so that's a difficult step and then you know with respect to back and forth between English Wikipedia and MD Wiki you know we're both under the same license material can be moved back and forth I don't think the English you know some of the English community don't want the material on MD Wiki yeah I think we're on time already sorry if it's just okay if it's okay moving on with the next session