 Then we have Ms. Rama Ayur, Director General WPP in the Foundation and Professor Preeti Rao, Department of Electrical Engineering, IIT Bombay. These two ladies will be talking about voice for social good with Happy Lingo. It's going to be a fantastic session for sure. I'll hand it over to Ms. Ayur. So we're here to look at how voice can really help be used for social good and introduce you to our app which is called Happy Lingo. It has been developed by IIT Bombay in collaboration with the Tata Institute for Technical Design and the WPP India Foundation. Could I just request you to have the slides up please? Just want to quickly walk you through the work that we do the WPP India Foundation. We work with the underserved youth of India to lead them all the way from educational outcomes to improved livelihood choices. Our model is quite deep. We work with the same child per period of seven years, all the way from the time that they're 11 years old to the time that they're about 18 years old. And in this time we work with them across three big challenges. One is how do we improve the school educational outcomes. The second one is how do we skill them and make them livelihood ready. And the third thing is we know that all this will not be possible unless we attest the deep rooted social norms. Our model is to develop proof of concept so that we can scale them with our clients and across geographies. In doing all this we do take a huge digital leap in two ways. One is in the fact that our children are digitally literate and they learn coding, they learn robotics, they learn 3D design. But more importantly in the way that we scale which is also digitally and we also look at impact. We spend disproportionately look at impact assessment. Our work is evidence based both in terms of decision making as well as in terms of the work that we actually do on ground. Next, please. Across the 23 interventions that we work in. In fact, like I mentioned, our kids learn a lot of 21st century skills. But one of the things that we have found which is most aspirational for them is spoken English and reading English. If you look at some of the challenges that we dealing with out here, it's at two levels. One is great appropriate reading. So typically a fifth grade student today is reading at a second grade level. The second area isn't dropouts. So one out of four children leaves school in the eighth grade without having the basic reading skills. This is an all India figure if you look at the communities that we work with these data points are even harsher. Next, please. So the question that we're all asking ourselves and for any of us who have kids I think we know that you know the biggest challenges are kids still reading and more importantly are they getting any better at it and are they getting more efficient at it. So we've looked at it at two levels. If you look at the government data, which is out from the NEP 2020, which talks about the huge challenge that we have ahead of us. We could also look at it as a business potential, but there are five craw children who've not yet attained the basic foundation literacy and it's becoming a huge priority in India. It's actually a challenge with all of us in the ecosystem, whether it's the policymakers, whether it's the teachers and the trainers, or whether it's a civil society or even corporates like us. Next, please. And I think for all of us on ground. One of the other challenges that we're dealing with is the actual implementation challenge. Do we do the assessment in class or do we do it digitally if you look at any of the classroom interventions on assessment. It's cumbersome, it takes anywhere from about two to three months to assess an entire school. It is also not standardized and most importantly it's not enjoyable for the kids. Next, please. To address a lot of these challenges we very happy to introduce our app Happy Lingo. It has been designed by the R&D division of IIT Bombay, which has been led by Professor Rao in collaboration with the Tata Center for Technology and Design and the WPP India Foundation. I want to talk you more about the app and what it does and how it works. I'm going to hand over to Professor Rao. Professor. Yeah, thank you very much, Rama for setting that context. So what I'm going to do is, you know, give you a kind of a description and outline the main features of the app and then talk about more about how we see where it's going. Can I have the next slide? Okay, so to describe the app, so we have a front end that's actually the part that facilitates the assessment itself by presenting text, first of all, allowing the user to select the content for the current assessment and then, you know, presenting the text and recording the oral reading of the child. At the back end, we have, you know, essentially what has been talked about a lot in this conference, the AIML voice, you know, processing technology that evaluates the recordings on models that we've built, which actually implement established reading bricks and prepares a report. Eventually, we have a dashboard that can display these reports and the reports revolve around these reading attributes that I just mentioned such as accuracy, fluency, expressiveness and confidence are all analyzed from the recording. And this can be done at multiple levels for a whole cohort or for an individual child across time and so on. And to give you an example, you know, just to illustrate, you know, how exactly it's used and what kinds of setting can we just move to the next slide. Okay, so this is a pilot we carried out in a school in northern Karnataka. So if you just play the video, we'll see how this child engages with the app there's a story being presented on the screen in the form of text and the child is reading it out the recording is getting made in the process and that's going to be available for analysis at the back end. So can you please play the video. The trolls are short and stiff children don't play with me. The next morning when that little tree woke up, it had big leaves just like the mango trees. Now while I'm happy, I said the buggy tree, but I great come along and a tea of all the leaves. Oh dear, I said the buggy tree. I wish I had gold leaves. Gods do not eat gold leaves. Yeah, thank you. Can we move to the next slide. Yeah, so as you saw, there is a number of, you know, comments and feedback that's possible on listening to this child read. So whatever your product does is essentially achieve all this automatically. So for example, some of the analytics that we produce are, you know, the distribution of these reading scores of reading skill on specific attributes like you see here. You know across let's say a whole class or you know a particular section in the school. We also have these individual tracking mechanisms so on the far right you see two plots on top which show how a child progresses across time over the course of months when these assessments are repeated. And with reference to the class performance. So we see there's a large gap between this individual child and the class performance in the beginning but the gap gradually narrows. So it's an example of some of the insights we get. And finally we have a pie chart there below which is kind of capturing very representational data about assuming that we have benchmarks in place about your snapshot of the performance across a very large cohort at any given time. So first thing we've been testing in urban very urban and rural Indian rural India schools and we find that in each, you know, setting, we do get something which administrators and education providers consider very valuable. Next slide please. So there is a case of, you know, obviously since we are doing this digitally we have access to a lot of byproducts of the analysis. And this is just an example of where we might be looking for at, let's say the you at vocabulary and familiarity and fluency with a certain vocabulary. So this just shows for example that in different settings such as rural and urban schools that are different words that trip students up. And this for example might be very valuable to someone who's designing the content or planning some kind of particular, you know, education intervention. Next please. Okay, so to kind of summarize some of the main, you know, kind of advantages of this kind of digital assessment. It's very clear that it's objective, because essentially based on computation and the computation itself has been trained, you know, on a large amount of field data. It's reliable in the sense it's repeatable consistent it's like anytime testing that's available and it's extremely consistent of course we digital and you know having this program in place and not depending on the availability let's say of teachers or evaluators and so on. It's scalable because it's digital. So that was a really big part of the design to make it kind of fun and engaging, you know, for the children to who are actually, you know, being assessed but all the same, you know, liking the experience that was a very important consideration in the design itself. So the comprehensive tool for spoken language, especially at this point for oral reading, which provides real time data. So we have a lot of data available and that can really be customized to the particular needs, you know, of the context and setting in which the impact assessment is being carried out. So our roadmap going forward. So we've done some testing over the past seven months is to build the app to assess students in order to increase its scope to more languages in particular we are looking at Hindi next. And to scale the app, you know, across India in terms of, you know, trying to pile it up for schools in India. And of course finally there's nothing really stopping us from going to schools in emerging markets and places where very similar literacy challenges exist. Finally, happy lingo, which is what the app is named has a potential to work with any existing technology to read general users who want to build their language skills. So I'd like to close on that note, you know, what we are looking forward to do and hand it over to Rama. Yeah, thank you. Thanks. I did notice there's a question in the Q&A professor. I think you might want to take it. Be quite happy. Maybe I can look at it. In case we run out of time. Okay, okay. Or should I answer that question? It asks about whether there are other available similar apps. Okay. Yeah, so I think just to give you a perspective, you know, this is part of a research project we did and it happens to be a very active area of research this whole, you know, idea of assessing spoken language where you're trying to listen critically to the way language is being used in this case, reading for let's say, you know, the accent, the fluency, and so on. So this is an area of research that there's something that we see a lot in, you know, speech as for off to self solutions, no, there are none. But like, you know, our previous panelists mentioned, you know, we have some of the framework and the machinery that's already been built by the likes of, you know, Google and Amazon and so on in terms of speech recognition engines. But it does require a lot of customization for this application because we need access to a lot of parameters other than just the words being spoken. So for example, we are analyzing a lot more parameters for fluency and expressiveness. So from the fact that we need, you know, certain other, you know, information that's extracted but that's not directly available from these engines. So that's the reason why we do not probably have already, you know, products that we can just pick and use for this application. Thank you for any of you who have any other comments or discussions we're very happy to connect with you offline. Thank you. Thank you. Thank you so much, Mr. Amaya. And of course, the Professor Preeti Rao for being with us. Thank you.