 I want to welcome everybody back to the Sunday afternoon. Great to be here, especially with all the hackathon attendees, but also, of course, all the speakers and guests of FOSS Asia and our judges here in the first row. Let's take you through this afternoon, running slightly behind, but please, let's get started. The teams, don't worry, will guide you through the process, so we've got a few assistants here, Mr. Guy and Ayana, and so we'll set you up before you come on stage. Our proud sponsor for this hackathon event is UNESCO, and so we're very glad to have them here again, and also part of the judges whom I'll introduce in a minute. This is the second year that we're running this UNESCO sponsored event. Please take your seats and kindly also move in for others that are still coming into the room. So this taking place here at the Lifelong Learning Institute again, and to give you some brief background to what's been going on these last few days. We kicked off the hackathon on the 15th of March and we ran this for really two days with a focus on building awesome apps, games and assistants to improve people's lives, and the focus has been to use FOSS Asia technology in the year of indigenous languages. So not an easy challenge, but certainly a very relevant topic, especially for Asia with its many different languages and different cultures. Can I ask you to be quiet please? So we've had 194 registered participants. The hackathon itself ran over two days, and we had 15 final submissions. So can I just ask for round of applause for all the submissions? Right, to take you through this, I'll just introduce the judges. Take you through the presentations, which we will have to keep tight, so I need you to be mindful of time. The judges will withdraw for consultation to evaluate the projects, and we'll have an award ceremony, and we'll have some prizes, and then a closing of the summit. Mario, if I can just ask you on stage to briefly... Leona, do you have a microphone? To briefly introduce the judges, seeing as you've been very much part of this process. Thank you very much. Yeah, so we have a very wide variety of judges who are really distinguished from some companies who we welcomed here, but also from the UNESCO and from the hacker community, so it's really awesome. I would like to introduce the judges here. The first judge here on the left is Mitch Altman, he's a maker and our open hardware hero, so big round of applause please. The second judge we see here, they're sorted by... I don't know, I think randomly, right? So we see here Davides Storti from the UNESCO. Davides is a supporter for many years of the First Asia Summit. The first one you joined in 2010, I found pictures and you looked exactly the same. So we're very glad to have you here with energy always supporting the First Asia Summit, so thank you for joining us as a judge, Davides Storti. Then the next judge we have also here was introduced by Davides Furious later and is from the UNESCO Bangkok office, Misako Ito. Misako is also a long-term judge for hackathons, kind of a hackathon senior now, doing a lot of hackathons in Vietnam and here also last year at the First Asia Summit. So, and the next judge that we have is the vice president and city of cloud and cognitive software at IBM. This year, cloud is a big topic also and we hope to see a lot of applications that deploy in the cloud, connect to the cloud, so we're very looking forward. The judge is Shankavi Selvadurai and then we have the founder of First Asia, I don't know, I hope you all took the opportunity to take a photo with Hon Phuk because she chose every day a nice, beautiful, traditional dress from Vietnam and yeah, if you didn't take a photo yet, please come. So it's Hon Phuk time. And then the next one is Matura Bikash Tripura. I'm sorry, I always have like difficulties here but you're helping us and this is also the topic of the Indigenous languages here that we have that we can also learn from Indigenous communities, Indigenous languages and so on. So you're a language expert and having this from this angle, Matura Bikash Tripura. These organizations, Zabarang. And the next also, Vinod Kumar is an APIC technical lead for Microsoft here responsible for partner teams and yeah, Vinod you also have a background like from India but also from Singapore so you bring also the different cultures and the world together. So this is perfect for this hackathon with having the different backgrounds. So thank you very much for joining us. Yeah, I'm also glad to be always part of this. I think no need words needed. Many of us have talked directly so it's me but actually I should be last. Michael Christen is in front of us. He is the founder of Susie AI which many also used in the hackathon. We had different workshops here. We have a lot of developers here from Susie AI so it's great to have the expertise, combined expertise here of the community with Michael on the board. Thank you very much. All right. So getting into the heart of the matter, the most challenging part is the judging criteria and just for complete transparency of course. The judges are evaluating technical implementation, the open data and open knowledge and the application thereof. We have two X, which is very important, scalability and replicability. We have inclusiveness and then of course the focus on indigenous languages. And last, there'll be an element of judging on the use of FOSS Asia technologies. So that really brings together the value of this FOSS Asia conference. This is the lineup so we've got 15 teams which is fantastic. It's going to be a fairly intense afternoon so I need the teams to be ready but also keep to their time. And they get four minutes to present and the teams should also make sure they balance between the presentation but also their demonstration. If they have something to demonstrate, the judges want to see that. They want to evaluate how much you've been able to deliver if that's been possible within your team and one minute for Q&A and we'll be very strict on this but we hope the judges have some good questions lined up for these teams. So teams just make note and we'll coordinate this. At the end we'll have an audience vote. So at the end of this we'll show this link again and you'll be able to access this URL and submit a vote and that vote will go towards a prize at the end. And so we'll get to the prizes later but there's going to be five prizes of which one is this audience vote prize. So make sure during that we access this link again. So make a note of this link now if you want but we'll assign the prizes later on. Mr. Guy, I think now we start off with the first team. Okay, that's what the website looks like. So please access this website. You'll find the button to vote there, vote on submission and that will allow you to vote as the session continues and at the end of the evaluation of all the projects. We'll then take a tally of these votes after the judges withdraw from the room and so we'll award these five prizes at the end of this FOSS Asia Summit and the closing. Okay, first team. Are you ready? LIL, is it? You'll have to bear with us. You're going to have quite a number of teams changing and with all the laptops and all the configurations we're about to have some of these set up challenges. Okay, over to LIL. Let's start the timer. Good afternoon everyone. My name is Arani from Thailand. Good afternoon. I'm Yoshita and I'm from India. I'm Sahitya from India. I'm Nilima from India. And I'm Saitan from India. And we are from ILL teams. As you know, a few days ago that indigenous languages are dangerous and some of them are very dangerous and some is still dangerous. So we are more focused on youth, more focused on youngsters who are good on technology and also good on technology. And then what the concept is we would like, we still raise awareness to help the children or the youngsters to come back to their own language. But however, we are not ready to prepare anything yet. But however, so we create this app to help the youngsters if they come back, so we are ready. This application basically consists of three levels for the youngsters, for the millennials to easily understand and approach it. The first level is basic level which consists of basic greetings of how to understand the language, to how to approach people from different languages. It only basic greetings, written scripts also for voice. And the second level. Basically, our second level consists of community. Here we connect the people who know the language to the people who are learning the language. It is basically type of a chat. You can see when I post, I don't know the language, when I post hello, the people who know the language, for example, we took the Kareen here, they will message the same thing in Kareen language. So in this way, the people with basics can learn medium level of language. The person is done with the two levels and when he attends good knowledge about the language, he can come into the third level which is the advanced level where in here the research work, the book, reading and all, when he wants to explore the language, he can read books and all. We have database containing all the book details and all. So that they can just look into it and connect with the scholars. We can connect them with the scholars so that they can learn more about the language. And as millennials are most used to smartphone, we are also planning to have Android application along with the web-based application. And we also plan to have games because that helps users understand more about the Korean culture in a wider context. Many people say that if technology is grown up, it means that it can lose our culture. It's easy. For us, we think that probably yes and no. If yes, it means that we are not ready for that. But for us, we think that we can say no because we can develop those things and we are ready for the change. As you can see, these are three levels. But we are planning for the future implementation. That is, we are planning to connect Suzy AI so that when I message the language, it should read the language because many people doesn't know the script. For the understandability, the Suzy AI will design in such a way that it can read the language and it can pronounce it properly so that people who are learning it can learn properly. Time's up. Okay. Good team. Let's see if we've got some questions. Run over calls. Any questions from the judges? You've got one minute for that. Thank you very much. You said you haven't been able to do a functioning thing, but I see on the submission there is actually quite a lot of code. So, can you explain more or less what you'll be able to do or how this is functioning? Yeah, we will be able to do it. Including the question, including the skills on Suzy AI so that this is also mentioned. There is a mention, I think, on the skills for Suzy, if I'm not mistaken. We are adding the Suzy AI voice for pronouncing the message. David, I saw code in the repository. It's just a basic code. We just wrote some code, but then we could finish it. If given some time, we'll definitely finish the project. Okay. Thank you team. Let's bring up the next team. The next team will be called Source Coppers Generation. And can the next team, Karana Rescue Translator please form up in front? Are you here? You want to connect? Yes, PC. Okay, so for those who perhaps popped into the 2-2 theater here on the second floor, we had all these teams sitting there in different groups, programming away for the last few days. So that's been quite a challenge. I mean, these volunteers take it upon themselves to participate in this hackathon and deliver something functional and prove a product in a short amount of time. This team is CrowdSourceCorpers. Okay. Over to CrowdSourceCorpers. So before he gets started with the slides, let me just introduce ourselves. I'm Akash. I'm from IoT Varunasi. And this is my team. I'm Amish Saktani, a proud open source contributor. I'm Rajay of Dube, open source contributor from Uttar Pradesh, India. I'm Shubham Gupta and I contribute to Suzy Project. I'm Sonia and open source contributor in Mozilla Firefox. Alright, so we do not get much time to prepare a fancy slide, so I'll just, you know, focus on me. So what we have built is called a CrowdSourced Corpus Generator, and people, those who don't know what a corpus is. A corpus is basically a list of millions of sentences and maybe their translations or maybe the POS tagging. So it can be anything, depending on what you want to use it for, alright? So why do we call it CrowdSourced? So obviously we are trying to get the translations from the crowd. But the problem is that a crowd cannot be trusted, right? So what our system is trying to solve is giving a metric whether a translation can be trusted or not. And of course, we have further improved on the features, so I'll get down to that. So our, it seems like we have some technical issues. Sure. Yeah. So our whole app, I mean the platform itself consists of two major parts. So one is the translation itself. So there's a platform which will collect the translations and there's another site to the platform which will be used to give a CSV data or maybe like any researcher who wishes to work with NLP applications or maybe machine translation. So the platform provides an easy to use corpus generator. So the problem we are trying to solve is basically crowdsourcing it. And is this actually needed? Because you see in case of indigenous languages there are not many translators who are working. And the people who know the language are either few in number or they are not accessible most of the time. So this platform is basically taking it to a remote level so that people who are kind of bilingual like knowing somewhat both the languages they can make use of this application to help the community. But again like I said crowd cannot be trusted. So a person who is somewhat sure about the answer and a person who is absolutely sure about the answer. So this is where I mean the algorithm that we have thought of kicks in. So let me rephrase what we have actually built. This is not a data collector. It's actually an intelligent data collector. Right? So let me show the math that we have used here. So this is basically a vector representation. This is just a simple vector representation. This is a plane in which we have vectors. So what we have done is we have say for a sentence we have 10 translations that have been given by 10 different users. So each translation is converted to a word vector and is represented on a vector space. So sorry So using that vector representation we generate a mean vector and the vector which is closest to the mean vector is selected as our best possible translation. But then again a standard deviation of the whole distribution is calculated and if the deviation is a bit on the lower side then we are I mean we can say that we are confident about the mean vector that we have generated. So again if the standard deviation is on the higher end we would say that we are not confident about that answer. So this is the main gist of our application and this is how it actually works. So this is the dashboard that we have built. So the translation part that I talked about so that is integrated and that can be integrated into many different platforms where all you need to do is basically take input from users. And this is the dashboard that a researcher might see. For now we have just tested it on a few number of samples but as you can see those are the English sentences and the indigenous counterparts I'm sorry the labels have been inverted it seems and the score that we have generated is being generated by the algorithm. I guess I have conveyed most of the algorithm itself so if you have any questions let me know. There is that easy to use button which you will click on. Questions? So yes I didn't get what technology are you using? Do you have a back end also? Yes yes yes we have a back end which will be used to communicate to any other platform which is trying to integrate our solution to this. So let's say we have Suzy and a Suzy I mean a Suzy user just wants to pitch in to this translation task. So he just gives in like a keyboard and we send him one particular sentence which he will have to translate and this data is sent back to our back end and the algorithm tries to figure out I mean tries to give that particular sentence that particular translation a score and which will be displayed here as you can see. So I would be interested if like let's say this sounds already pretty complicated I mean like the idea is clear but maybe it's the implementation is always like right like a lot of details to it but could this be also used for example for roles like let's say we have like a role which is based on trustability so a person can have more more rights if they reach a certain role for example have a number of edits have verified an account things like that and also voted trustworthy by a lot of other trustworthy people. Yeah I understand your question so that can be a possibility so if a person is a valid translator he is a qualified translator maybe his contribution could be given a more weightage as compared to the other translators like for me I don't know German and if I try to translate a German sentence maybe my translation would be given a less weightage as compared to you right so that can be a possibility we haven't done that here but yes it's an extension which can be easily integrated into this we are not using a set we are trying to generate a data set this is crowd sourcing the generation itself so we can either take yeah yeah so crowd sourcing the data generation that is our aim right so there are already available I mean openly available corpus so we can just strip away the English part and we can use that English part as the set from which the user will be given the sentence or indigenous translation right okay round of applause for this team thank you okay so to team crowd source corpus I think a good recovery from your technical hiccups but teams please be aware that you need to be prepped and ready to plug in and have your slides come up so we don't have too much time to lose right so judges I think comprehensive presentation on a fairly technical subject matter but also good application of of Susie next we have Kerala rescue translators Kerala rescue okay if not we will have to skip popular online indigenous crafts okay popular online indigenous crafts okay team you're gonna get four minutes for your name and we're gonna have the timer on the left hand side here okay next team will be twig genius are you here twig genius hello good afternoon my name is Panom Tano I belong to Korean indigenous group I work with I work with indigenous media network in Thailand so hello my name is Ishaan she's Ngarika and he's we are from India yes so I'll just briefly introduce you to the indigenous people and indigenous people to access to this kind of platform in this app we allow indigenous people to publish their own community story and sell their own products from their own communities and we allow the youth indigenous youth who are living or studying in the city to access this kind of app and bring a product from their parents to post for selling on this kind of app so this app if this app builds it will be help indigenous people to revive just not only their language but their culture their tradition knowledge in making this kind of product and relating to natural resource management and ecosystem as well so this is actually not a shop shopping site but it's going to help the charity and NGOs that are being that are available in the indigenous communities to grow like if if he's a customer he's going to come and search for the communities looking to contribute so once he likes some of the communities like suppose there are some cancer based NGOs some NGOs are working for the women and much more communities are there so once he select any of the communities he will be able to contribute and he will be able to purchase any of the products so apart from this we have we also have got premium premium membership that what it does is you need to pay like only $5 per month if you pay that money you will become a premium member and you will be allowed to attend the event that is going to take place in the native places of that community or NGO also apart from this you will get the products for free monthly like that are being made by any members or the students of that community so the basic idea behind this was that suppose a person is from that we will be able to recognize each other or we will be able to talk to each other but when we come in contact with some products like if you buy some products we like some products we tell our friends that these products we get from this place and they they all build it so this is going to help the community so this is the video of the app so these are the communities if you select if you first login so these are the communities so you will get the information regarding the communities and once you select in your favorite they will be available in your checklist and you will be able to buy the products then yes this is so Harshit has a prototype so if any of the judges wants to see you can check it yeah so thank you you guys are good on time well done team very smooth let's have the judges ask open data open knowledge yes all of the data are open data that provided by indigenous people themselves how how do you get the data how do you get how we get the data right because we work with indigenous network for example I work with indigenous people that work with indigenous people in 14th country in Asia and also work with from 48 organizations and we have a group working with indigenous people in five countries so we can just maximize this kind of network alliance that we have to introduce the app and bring information to them further questions from the judges time for one more just on the security side of it how are you storing data how are you enabling that into database escalate database and we how are you thinking about security and storing all of it till later we can use firebase because currently we have limited time so I just hard coded okay so anyone of you wants to see the prototype if you have okay you have shown your prototype okay because we are tied for time I think team you're up so thank you very much okay so our popular online indigenous crafts sorry team was certainly aware of the complexities of indigenous languages and also the the number of challenges of sourcing the data I think that's the biggest hurdle in any machine learning is getting access to the data in order to train the models so going down to grassroots level is definitely one way of doing that this team is tweet generous right okay so next team up is tweet generous here's your microphone can we so our team is tweet generous tweet generous is basically a Twitter tool for indigenous people currently like the government does not have any system in case of disasters to like take data to give data or take data from Twitter to provide help to the indigenous people so we have developed a which will get data from Twitter and like using the hashtags we will use the hashtags as the input once you put a hashtag you can choose a language in which you want the hashtag once the hashtag has been given given our return the tweets with the hashtags and other tweets which have similar hashtags with this the government can find out where the natural calamity calamity is like worse and send help to the indigenous people this is for helping the rural people as well like they can check which if any natural disaster is headed their way I can move to safe shelters and this is basically what we do a few apps for sessia like the pslabs Android app is a really great app it is very good for students to learn science but the app is currently only in English so like it cannot be used by everyone we have localised the pslabs app in Nepali language so that people who speak Nepali and are not really well versed with English can also use the app I shall show the screenshots of the app and our site as soon as the English is resolved please bear with us I would like to discuss the web app in the web app we are taking two inputs first would be the hashtag of the specific event that he wants to that the government or any NGO wants to search next input will be the language that he wants to search any indigenous language can be added but at the moment we are using Nepali as an indigenous language when the user will click the submit button we are using the idea of millix generator that is another for sessia travis will and the travis will run a python script and will give us the data that will be shown in graphical form that will be discussed by my colleague rahul hello yeah this will be the front end part here we will pass the query like I have already taken this is based on the Nepal earthquake so here I will be using Nepali language so it will face the data we have already collected from the twitter so this will face and how it works first of all we have used looklack to collect all the tweets and we used some python script to extract the hashtags from the tweets and we processed all the tweets to reduce the noise so that the precision of the system can be improved so you can follow the implementation like we have used the library and we have imported the file which we collected from the twitter and after pre-processing are all involved the pre-processing part and after that I have extracted we have extracted the hashtag using some regular expressions and we have plotted the hashtags here you can see some hashtags like home government can use this data to provide the relief very effectively like in the tweets the hashtag home is used frequently that means there may be chances that people can use to build some more refugee shelters so that they can provide some pre pre-data can be obtained and here you can see some reverse names that means in those situations these reverse have badly effected so government can use this data to extract or remove those people from those areas this can be previously can be done before the before the high high casualties and here you can see some state some districts names are also visible like Setupati and Hamrakura so government can use this data like use these states to provide the priorities so that government can provide the service first to these cities then to other respective cities okay just a round of applause I think this is particularly challenging when you're under pressure and stressed to get up and running over to the judges this is already on stored data that you're doing it or is it on this is the BS Labs app we have created translations for the app and it is very functional it works if you want I can show you a live demonstration on my mobile it's on real time data or is it on stored data BS Labs is the existing app so that data can be viewed in Nepali as well currently it is only in English demonstration that's fine okay thank you very much okay team right unfortunately but challenging to present without slides so let's hope we can address these challenges you are good afternoon everybody yes my name Samin Ngaj and I'm belong to Pundong Indigenous and I'm from Cambodia and I'm working for the Indigenous HHS and I'm happy to be here today and my team will present our home of Indigenous peoples and my friend will do that yes good afternoon everybody my name is Don I'm from Laos and we both are Indigenous from Mekong region and our web developer are not here with us so basically we are working on the website by ourselves so we will show the website so the idea of developing the website is to preserve and promote Indigenous language through animation and also use our culture by learning English through the culture through sport art yeah then you can talk about the vision okay this is our vision and mission that we create the home of Indigenous people and the vision we want to protect the cultures social and traditional knowledge our traditional system of Indigenous people collect and share the traditional knowledge about Indigenous groups because our website is Indigenous in the Mekong but also we can look for the who are working on the Indigenous issue and also we empower Indigenous people especially we promote the rise of Indigenous people to land without land without right and without culture too okay and just next slide and how we can do this work because we have the specific plan to design our work and also we will build the database like we have cultural history for each group like art photography record just something like this okay next yeah here is the way forward that we will work like training the young Indigenous people in technologies and also teach how they promote the cultures in the website and help Indigenous people to communicate with the international community through using a dictionary and translation Google translation as well this is kind of thing that we think it's important for us Indigenous people to do kind of work yes this is our website yeah you can click yes here is our website I hope everybody will enjoy can you show us the website briefly you've got a bit of time left yeah so if you go to the home and then you will see this is the idea that we want to post for example this one yeah this one is Bruno Indigenous from Cambodia and this from Camus yeah like this and then when you click to each Indigenous and then you see the language so you click on like let's learn Bruno language and the culture let's learn Bruno language and sport like something like that yeah okay thank you very much okay round of applause so the judge has some questions to the website of cultural artefacts and learning any questions from the panel so sorry how do how are you going to fit in the site with the data I didn't get it properly so how do you maintain the data I mean okay thank you for the question how we can maintain it so first we will do research and then we also will allow Indigenous from the region to contribute the information especially on art culture those information that we still don't have much and also their art especially the drawing painting related to the culture yeah and then we will feed the information on this website and basically only two of us right now as a team to maintain this website so as you can see we haven't done much because our skill are very limited and we don't have web developer to support us of course we are not just only two but we have indigenous people back we have big network so I think to maintain this website we will collect all information to post in this and we will launching this kind of work that we have done here this is the thing that we will do next how to maintain this kind of work thank you okay good thank you right I guess maintaining a website like this as UNESCO would know is a big challenge getting all the data organized and sort all this content is a daunting task but very good presentation team okay next team up just getting set up this is the Guageland is that right Guageland Guageland okay so we too have challenges with languages Guageland there we go okay over to you team four minutes starting now hello everyone we are team Guageland so this is our product Guageland a land for indigenous languages okay so next page so now I briefly talk about our design idea it's all about exploring all the endangered languages in the world household land leveling up system and rewards we also use the sushi AI part and quiz and findings also we have friend circle sharing these are our main features so here are our leveling up system you can see the houses on your own land okay so for the for the business model we have the travel agency and the endangered language project involved in our project in our application okay so the technical model we use something wrong with the technical for the business model for the technical details we have Reagis sushi AI and AWS DynamoDB IBM switch to text HTML5 and CSS3 progressive web app and Node.js okay so looking forward we have some main features okay so let's come to that yeah here I'm going to show you a deployed version of our production it's quite cool I think yeah here we go yeah okay so yeah on our phone you can see a panda yeah actually we designed this logo by ourselves because we the panda and the endangered language we hope our app can also help indigenous language as popular as panda yeah so when you log in you can be directed to a discovery page so every day we will post some random news and article about the endangered indigenous language and here is like I love you in some kind of language yeah and we also integrate the sushi in our app so let's say hi to the sushi yeah tell me a joke right and tell me what is a greek land yeah yeah we will save the language yeah okay and we also have an interactive earth so you can go explore the earth to find indigenous language so let's explore Southeast Asia and let's go to the last so this is one of the endangered indigenous language from Laos and you can see the culture and some information about it yeah and here is the rewarding system so you can see your points so with the point you can level up and every time when you level up we will try to find a sponsor and you will actually donate like hundred dollars to the local language and here you can see the future house you can have with your reward points yeah it's quite interesting right okay yeah and here is your profile you can see the organization like in the future we are going to put more organizations organization like we are going into have the partnership with so right now we only put us but in the future this is how you are going to earn your points we have some quiz about the language so let's try the quiz and after you finish the quiz we will give you the points and then you can build more house and donate to the local organization yeah that's all about our app okay team thumbs up do you have any question yeah yeah and we have actually deployed to production it's not just local development any other questions from the judges please we want to test these teams so for a database we use the DynamoDB AWS DynamoDB it's a no secret database I think this is more flexible for our user data and the article and the language yeah I think it's very good for our progressive web basically also you see our application is actually a website so the accessibility in mind basically we can extend to different devices whether it's a mobile device it's an iPhone or Android or anything else so we're actually reaching out to a wide range of users who will benefit from our app and the indigenous languages will also benefit from our rewarding system okay thanks for that team well done okay great quite a comprehensive and functional demonstration with some interesting technologies at play I was also curious what data comes from because of course that needs to be pulled from somewhere and displayed accordingly and also hopefully be accurate right so onto the next team that we've got lined up is that vanilla so vanilla is up for the judges vanilla and I think we're good to display is that right right Dan let's start with the timer four minutes hello everyone we are vernacular so first of all what is vernacular an application for the people of younger generation to get involved with the indigenous languages of the world so why the younger generation why not the older generation first of all so the older generation they know that they know the indigenous languages quite a bit but the thing is due to the globalization due to people going out of their local communities people are going towards English languages are a bit diminishing so we want to involve the younger people using application based on Susie and that is not just an application that is also a smart speaker based application that is also a voice application also so we are making a quiz game out of Susie that will help us involve more younger people into more local so we have used Susie and low class so we have to generate a list of sweet a suite of games based on the indigenous languages that the user selects so like I wanted to play a game in Kree and we have created an option selection game in Kree as soon as the game is finished the high score is posted to Twitter this makes a community through the low class and creates the community and connects the native speakers to the people who are still learning so yeah building an application building community out of it that's our main aim that's our main into this application so what is the achievement we have not just built an application we have also built a skill that can be easily used by the native that mobile speaker of Susie and even small child can play this game this is so easy we can we mostly want to include more folk loads and everything that people have like you remember sitting on your grand parents lab when you were younger and listening to the folk loads they used to say Susie is like a family member you can easily access these things you can easily learn more about the indigenous languages for sure so obviously main thing we are making a better community out of it made a world out of it so that's it from our side thank you any questions please okay excellent so we have created an iOS app with a few web views and we have used the low class API the Susie the chat.susi.ai app and the Susie skills data to data custom skills and the games are generated from native HTML5 and canvases and iframes another question from the judges about the gamification interesting concept how right now for the indigenous languages there's not much open API or open data available so we had to stream the data mostly and we are getting in a database where we are making a different list of words that will be used that game is based on that like specific word what is like what is an I increase that specific word it is so we are making a database out of it we are making the translations out of it that's how we are making the database okay we're good so round of applause good thank you team vernacular it's an interesting take on mobile app gamification like the previous team making it a bit interactive and using Susie AI so it's great to see these projects of first Asia technology also okay next team is up and ready we team manang right over to you okay good afternoon we're team manang which means friends in kachin whereas we're very lucky to have a kachin expert join our team so so fundamentally the problem we're focused on is in these indigenous communities basically the children they have to learn the national language which doesn't give them that much time to learn the indigenous languages so how can the previous generation transmit these reading and writing skills to the next generation so basically we want to build a game that can help the kids learn to read and write in the indigenous language so fundamentally the most important technical problem is a lack of data so we don't have a lot of data in the indigenous languages so the problem we tackle is how to collect this data okay so here you can see puji our expert is actually recording his own voice samples so we deployed a website where they can record their voice samples in indigenous language okay so the AI problem we want to tackle is how to train a speech recognizer for the kachin language so there are many ASR systems many languages probably none for kachin right so what we want to do is we don't have a lot of data in kachin but could we use the English data sets and kind of fine tune for the kachin language the solution we came up with is that they use different alphabet systems English and kachin but if we translate everything to IPA the international phonetic alphabet then maybe we can transfer some of that data set okay so we actually were able to do some training during the hackathon so we did this on Saturday so the data set is called timet it's an open source data set about 6,000 utterances 400 speakers in English so the first step you transcribe it to IPA then we use PyTorch and some open source to train RNN and just to show we actually ran it it took about two and a half hours on GPU finally what we would actually do is we would take the kachin samples and we would try to fine tune the language model and use that language model to create an ASR and kachin so good okay so just briefly about how we prepared the data so first we got our expert Fuji who speaks kachin and basically he gave us a bunch of words kachin words and we first have to try to translate it to a few names in CMU which is Carnegie Mellon University's Phenetics Phenetic Alphabet because that's easier to type in and then thereafter we have a dictionary to translate it back to IP so as you can see there are words like Godin means football Manang which is our team name it means friends and Singapore it's not Singapore it's that's the right way to say it in kachin and there's Tom I and so on okay then here's the demo so the demo is basically to so the game is like you have to say this word in kachin so assuming that I don't know any kachin I just go ahead and say something so sorry I didn't say anything so I should try the hint button it says so when I try the hint button it says you should say Godin so Godin means football I would say then okay I say so it's actually like in the end it'll pronounce what Fuji had said but then Godin sorry okay very interesting very interesting topic is the code uploaded in GitHub? I think we have a GitHub we can share the link for yes for the questions to kachin and what was supposed to happen after you say the word so in the demonstration but would I expect to see I think the point is if they can pronounce the word in kachin okay then I'll give them a point but if they can't they'll say okay here's how you pronounce it and then we try to grade their performance on how well they say the word so I can train the speaker yeah so using the AI to grade the the speaker's performance yes where's the part of it on this when you said when you're training it okay so yeah we trained a model for speech recognition and we still need to fine tune so actually we were able to collect four different speakers speech samples two from Myanmar and one in Thailand but obviously we probably need a little bit more data, a little bit more time to kind of fine tune the model any last questions from the judges okay I think the judges are happy I think our time's up thank you team okay I think we're looking forward to hearing some kachin always novel to have everybody from different cultures but different backgrounds too and we can share that in the teams we have both students friends technical people non-technical people and even I guess family members participate so it's really bringing people together to come to a hackathon and spend 48 hours together turning out a product in a rough prototype over to the next team which let's make sure I've got information here is this team Simru? good afternoon everyone we'd like to present to you our project hazin which I'd like to go into lengths about why we named this if time permits at the end so I'd like to begin by introducing the team members my name is Anna my name is Nick I'm Lam I'm Isu Robyama so first we'd like to talk about the situation at hand as you all know the indigenous languages across the world are decreasing rapidly as that's why UNESCO has declared 2019 a year of indigenous people and so for example we have a team member with us Su who's an indigenous person from Myanmar specifically the western part of Myanmar and as of now there are only 1.5 million speakers of her language which is Arakanese and this may seem like a large number but compared to 360 million speakers of English this is a drastic difference so in order to preserve their language languages such as Su's we came up with an initiative in order to help fuel and preserve the languages and that is through our gaming application which I will hand to my team members Nick and Lam to talk more in detail about so we developed the idea and the problems that a lot of our competition actually faces is that when dealing with getting data for speech we don't really have a lot of that for indigenous languages so we decided to gamify the collection of speech data for the use of other applications and we did this through adapting Flappy Bird and being inspired by that not connected chrome dinosaur game and got so the idea is that we have you have to speak to get the bird to go over obstacles and the obstacles and different objects that prompt the player to say different things in their indigenous language and we record that and play it back and associate the obstacle like the image to what was said so I'll demonstrate here right so I'll try to turn down the volume a little bit so we don't blast I'm not sure if this is actually connected to speakers at all but yeah so the idea is these buttons click on the buttons on the side and the audio that are recorded is actually played back and then in the future we would extend this so that we would send these audio files to a remote database along with the associated images so that they could be further studied and used for machine learning models and then we would the idea is that we made this game very easy to expand you can add in just you just provide the path within the code and you can just expand it to get whatever images you want and you can apply it to any given language it makes no assumption on whether or not the language actually has a written component so in addition to that we're saying that our target audience for this app would be the indigenous people and it's a good way to help I guess taking their curiosity given a game to serve it for future data sets for maybe future linguistics or technologists who hope to use this data to fat fuel other definitions in the future and although this game right now seems like you can just you can just say something forever and then it'll stay up there we can have obstacles at the very top so then you'll have to not speak and then whisper at some point in order to avoid the obstacles so it gets more complex but this is our basic rough draft today thanks team nice over to the judges for one minute of questioning the technology what are you using and how are you storing now today I guess so we did we ended up not using any of the fosatio technology and we were going just to try using the previously developed flat paper game and that didn't cooperate very well with the microphone so we just used plain JavaScript we took a couple of open source projects already to get the basic game physics Mario you have a question do we have time for one more question if the judges have nope we're good okay Tazin well done good presentation right so we continue on next team up is called NYJC curious about the name yes hi good afternoon judges so our name is basically NYJC as you can see it's the name so we decide to take on its name and hope for good luck so our project is very simple and it's purpose I guess so it's up and running now so you can access the URL link if you would like to so yeah it's a very basic web page that we web app that we use to promote and we raise awareness of indigenous languages so as you can see we have a Y page which basically show the reasons the statistics and what you can learn about the indigenous languages as you can see there's a learn more button because basically we're not experts in the languages ourselves so we have a learn more button that link to a web page where we deem it's possible for the audiences to get a better overview of the matter so yeah okay so we also have a language map which we painstakingly input the data of the indigenous languages yes so yeah so if you click on the nation or the region on the map it will show you what kind of languages there are in that area so yeah we implemented part of but as you can see we have not covered the map so that's kind of a point of improvement that we can implement in the future so we have also a what page so basically the what page is where you can learn so here you can so it's a demo so I think I should pass the mic to my friend right here so he can so it's so we use Susie AI as the base because we think that languages are very dry we are students so we have no experience with languages so it's a very dry topic and we think that if you interact with a chatbot that's better so basically we passed some skills to it so it's not the basic Susie AI that you find so we added some language skill to it so when you ask a question like that so it will pass back an answer of language that we've already implemented so it can translate from Ainu to English Ainu is a language of indigenous people of the northern part of Japan so it's Uncle Ainu is Akapo as it seems so we also implement the random word if you want to know more about the word because we have a word list but apparently no one's going to browse it so if you want a random word you can actually ask for the random word yes we also implement the fun fact so if you want to know like maybe a random fact just some small fact about the indigenous languages you can get it from Susie AI also so yes also as you converse with your chatbot sometimes you will like sometimes your curing will contain some like interesting word your time is up team okay well done an interesting demonstration of your different language aspects one minute for the judges thank you I was wondering about the the language how do you implement the language so as of now we only have Ainu as our language so as you can see we have a language tab there we intend to put more but of course of time constraint we only have one language as of now so we can translate from English to Ainu and back but yes but how is it done oh I see it's on the table translation basically the basically the language data is stored in a way in short it's a 2D array so you could just basically extend one column to add for another language so yeah that's a manual thing at this point we're just using it as a key value and then just extending it that's what it's okay judges I think that's all time we have our team and YJC and YJC and YJC so thank you okay yes there we go so as we progress through this we only have three teams left if they're all here I'm very encouraged by all the teams that have come up so far let's have you come up and we are Gadong yes we are Gadong okay team Gadong we're gonna give you four minutes good afternoon everyone I'm La from Laos actually our team is three plus I will talk I will share a little bit about the reason why we came up this idea in Laos we have the problem that especially Kamo people in the community they don't speak Laos so when they go to school that's a problem that it's hard for them to study and in the same time in the city Kamo people they can't speak Kamo so that's why I created the app so I will hand out to my friend to talk more about this app thank you Laos yes so now I want you to meet Noi she is a Kamo a female and she gets the opportunity to actually learn English through our app at the same time she helps us build the English Kamo dictionary to your right you can see Jane she's a tourist now in Laos and she finds it difficult to interpret the language so our app is actually going to help her do the same so let's highlight our problem statement inadequacy of online resources for indigenous languages Gadong so Gadong is basically a game we have designed to address the inadequacy of these online resources for Kamo which is an indigenous language it helps provide translation facilities as well as a dictionary now we will have an online a demo okay so I'll do the first part and the first part is so we're authenticating with any social media now it's Facebook and now we have two options one is play a game or two is look at the lexicon so if we're going to play the game and you get a word and you get the picture and you also get the lao on the right and because Kamo doesn't have official text they're going to type in lao and so I have some just imagine a local was using it they would type this and play the game and then it would go one by one and this is basically to give them incentives they get points for every question and there's a leader board that they can check where they're on that leader board so the more people play our objective is being reached which is getting Kamo words which is an online right now and he's going to talk about the lexicon okay so basically the words that we collect from that game are the ones that we're going to use to populate a sort of online dictionary that's going to be open source and people from anywhere can use it because we see that a lot of other developers and even people at UNESCO who are interested in working with indigenous languages and we think this would be a great tool for them so in addition we've also implemented a translate feature here so the words that we collected to build the database can also be used to translate by tourists or other people wanting to so technically if you go to Google translate you can't find this because all you can do is English to Lao Lao to English this is giving you English to Kamo and so it's we got this idea basically because we were speaking to Lao and she said that her indigenous language doesn't have an official script and this is supposed to be a fun way that you gather data that doesn't exist and the dictionary is supposed to build and grow and yes because of children they can guess by pictures so that's why pictures are easy for children to understand too and again I want to talk more about Kadong why this app means Kadong or call Kadong because Kadong is Kamo language that that's very popular for Kamo you can see every year we have Kamo new year so people share story and song around the Kadong it's very fun so it really represents Kamo so that's why it's Kadong well done team okay nicely presented okay we've got one minute for the so we're actually using an online MongoDB storage and we're putting it there for now do you have any more questions? this is a progressive web app so it's supposed to be a website that functions like an app we're just showing it here on the website also that it's cross-platform and built-in React using Wordbox where are you getting the English word from which to translate from? okay so essentially we kind of used a random word generator which gave us a lot of words and then we actually put it through Python doing parts of speech tagging to get like words we thought would be useful to translate and then we use that list to get the images and so open source okay great team well done team gadong okay so for team gadong and some really indigenous languages well put together with a small team to make it so functional and demonstratable I think also important is the consideration for web technologies and mobile technology in the context of these emerging markets mobile first making these things mobile accessible the team that we're going to bring up now is a speaker from earlier from Simru so judges just to make a note we're going to Simru from earlier and you get four minutes okay okay first of all it's not pronounced Simru it's Welsh and it's pronounced as Camry yes Camry weird isn't it now let's talk about Welsh well Welsh is according to the UNESCO languages that is kind of endangered is a vulnerable so it could go in danger but it could also not go in danger but yeah so it's kind of going out of trend so Welsh is a very interesting language so I have a solution for the promotion of Welsh language to make sure it does not go extinct I got a and that's Simru and Biff it's pronounced like that it means Liflong Wales it's a national kind of thing for Wales so my game is Simru it's a Welsh English soup of many beauty and yeah it's me Unity it's a game mate so I'm going to present my project it's going to promote the indigenous languages so yeah Simru it's just you know this kind of game anyone can make this play around so I want to play so yeah if you're a kid it's like okay man I'm going to use my arrow keys to move okay I'm going to go left this ball is Wales you know you can kind of get the picture no need words much so yeah the ball now it's Wales on top of it and the kid you know he just looks at this and it's like okay I need go to the Simru I don't know but wait where do I go now yeah I got these two floating over me and oh I see two words and Biff long live Wales that's I infer yeah I have a boss battle yeah go Wales Simru and Biff UK damn imperialist out at Wales to defeat the UK one bullet it dies UK Wales wins Simru and Biff the end promotion of Wales now why would this promote the Welsh language cause it's you know jokes people get it no like words it's very easy to remember it's funny it's memorable you don't need to teach them an entire language to remember some language you teach them words like words that by word by word so a game as a medium is like the best way to teach like people languages I mean it's memorable I mean you all I mean English speakers you just pronounce this as Simru but you know now you know it's what is it again Camry yeah Camry and Biff not broth so the end I end my presentation now do you have other people in a team or is it just you doing this yeah just me so how did you build the game so this is made with Unity Unity engine it's like many people use Unity engine cause number one is free so anyone can do it anyone can like make a game on it so like yeah I mean people you got a statement to make or something to promote whatever you just whatever engine you want to promote your statements so you want to promote your language just make something funny make something memorable so yeah so anyone can do it even me a guy is really lazy I mean it's kind of rough it's right this is in AAA this is something someone would pay 60 dollars for why did you pick weights cause I read many political comics and Wales is like portrayed very hilariously like like Wales is like you know like the guard dog of the British like the cheetah is like an attack dog according to like some of the comics and like yeah it's kind of the areas yeah it's behaviour if I say any further I might get in trouble so yeah okay so thank you so much Kaimo for yeah entertaining so something more different and I guess experimental but making use of the unity framework for this rapid R&D and demonstration so we're down to the last I think it's the last team Indy Genolope I just want to ask is the is the last team growing tourism here are they in the audience you are which team are you you're Indy Genopoly right okay so this will be the last team for today give it to them okay let's see to get this slide deck up here so we're just going to switch laptops make sure that comes up okay team I think for the sake of time and making sure you get your four minutes okay let's start all right so I hope that comes up at some point of time if not that's fine all right so good afternoon everyone my name is Krithika Adrian Anjali Sophia we've been working on this project together over the last three days well so as you can see we are a mixed race family my husband comes from Romania and my kids grow up with a strong sense of culture from both sides of the world and when we were trying to teach our kids Romanian we noticed something very interesting we noticed that there was there was a very high overlap between the words in Romanian and the words in many of the Indian languages for example in Romanian there is a word called dushman and the Indians over here would know that the word dushman means enemy in Hindi at the same way there is a word in Romanian called prieten and prieten means friend and the root word for that is priya which means somebody whom I love or a loved one and we were wondering what actually happened here why is there such a similarity between the two languages and in fact there are 300 words in the Romanian vocabulary that come from Indian languages and we found out that this is actually because of the Roma people next slide please so the Roma people are basically Gypsies and they travel from India and they are now living in many parts of Europe they have a culture they are an indigenous group of people and they are responsible for some of the similarities between the languages that my husband and I share and which is which is important to the roots of our kids however we all know that the Gypsy people because of the Gypsy like lifestyle they are persecuted so next slide please okay so this is what we have done to help so we made a monopoly based game which has many different tiles and each of these tiles have a light sensor so you can detect which player is on which tile so there's an Arduino over here which controls all those light sensors this information is passed to the app and now okay so this is the app it's part one there are two parts for this and we are going to chat with Susie we'll just click that button and so yeah the thing you were sometimes Susie wasn't really behaving with us we were asking it whether it could tell us more about Romani which is what this feature of the app is mainly used for the second part of the app would be the game board so and in the game board as my sister said players move along and it's connected to the web server so when these pieces they allow different amounts of light to pass through so the light sensors can detect them they move over will give different numbers to the web server which this app will then take so for example if a person lands on a forest tile they'll be able to buy it and for this they'll have to use points and these points come from the learning portion of this game where if they land on the learn tile they'll be asked a question and if they answer that question correctly they'll be able to get a hundred points this is the technology we have used Tunkable for the app Python for the backend and the flow control Arduino for the game board and all the messy stuff back here Susie AI for the chatbot for the users to learn more and Python anywhere for the web server okay this is a very short implementation what we have done in two days we can extend it by adding a lot more information personalized Susie AI skills and implementing a rich text format that can allow people from very different backgrounds to contribute with information based on their own indigenous languages children do not have ingrained prejudices therefore we have made this a game to expose them to different cultures early on in their life so that they can create a cultural bridge okay thank you well done and we're also good for time so one more minute for the questions judges any questions to monopoly no questions but I applaud the kids for getting involved okay any last one more from Tavide yeah unfortunately we haven't seen much of the demo so it actually works pass the microphone so let's see if we can just play the audio for that okay let's just see if we can bring up this video for the audio out so we think the audio of this is going into the HDMI so yeah this is the first part of the video again with Susie AI my mom is asking me a few questions but I think the volume is dead on this laptop does it work okay I think we're up for time so team thank you for trying and really putting this together right so very interesting to learn about the spread of languages and going from the subcontinent of India and finding words that end up in a different part of the world and I think that all speaks towards this this initiative of UNESCO to give some attention to these indigenous languages so where we are at now is at a very important juncture of this afternoon where we're going to be collecting feedback from the audience about your popular audience vote and we will let the judges adjourn for their votes just to take you briefly through it there will be five different prizes and we'll recap that in a minute I'd like you please to navigate to this website and cast your votes and the judges I'll let the judges adjourn first and withdraw for their consultation in the meantime for the rest of the audience we have some snacks outside so we can have a quick break and we look to regroup here shortly so take it in 10 minutes and try to be back here shortly after 5pm