 Today our topic is localization for Story Weaver and we'll be talking about approaches and the challenges that the Story Weaver team at Tathambukh has encountered. And we have for the first time two guests together who are going to share the story with us, Suman and Amna. So can I ask Suman first to introduce yourself and then Amna? Sure. Actually full name is Sumanah Manjunath and people started calling me Suman which I also like so please continue to call me Suman. Currently I am playing the role of implementation head in Prathambukhs. Okay, so otherwise I have been playing this implementation head role in prime focus technologies previously and then long career in Vipro around 17 years and I was also a scientist in Israel. So that's the quick introduction. Wow. Amna. Hi, this is Amna here. I'm a marketing communications post grad, started my career in advertising and I think ended up in the social development sector as a rebound, you know, the rebound affair. And it's been a story, a joint story we were right at the Mason stage which was around January 2016. I am the content partnerships manager on the platform. So my role involves nurturing the global communities for content and building content and local language libraries. So we work with various stakeholders from publishers, authors, translators to literacy organizations, universities and also indigenous community groups who are trying to archive and safeguard the language via stories for children. All right. Amazing. Welcome both of you. Thanks for taking out your time for this session. So this particular session, I expect it to be very relevant for people who are thinking about making your sites or your products, be it large or small, in multiple different languages. And specifically for people who work in the content side of things as well as people who work in the tech side of things, so content and developers, development teams. This session is going to be very relevant for both these groups. A few instructions are, this particular session is going to be about 20, 25 minutes of the speakers going through a set of slides that they have prepared. And after that we'll move on to Q&A. We won't take on a lot of questions in between the slides when they're speaking through their prepared talk. And after that, about 30 minutes down the line, we are going to start taking questions. But in case you have questions and you are on Zoom, feel free to use the Q&A panel to put in your questions. I'll be taking it at the end. And if you are watching us on YouTube, feel free to put in questions there because I'm told that those questions will also be forwarded to us. So that's all that I have as initial starting instructions. Over to Suman and Amna. Please go through your presentation and then we'll go for more questions. Thank you, Saurik. Thank you. So I'll start off this presentation. We thought about our journey of localizing StoryWeaver. It's been an up and down roller coaster ride as far as localizing the platform was concerned. But before we dive into what were the challenges and approaches and now the kind of processes that we follow and where we want to go as far as the dream process is concerned, we thought it would be a good idea to start off by setting a context to what StoryWeaver actually is. StoryWeaver basically is a repository or digital repository of children's picture books, multilingual picture books. They are all freely accessible on the platform to read, to download and also to create and translate. So basically we have two universes. You could come and consume stories on the platform by reading them online or by downloading them for offline environments. Or you could also come and contribute to the stories on the platform. You could create your own stories or you could also translate the stories on the platform into languages that you are versatile in. The idea was to bring together the whole educator librarian spaces because we knew that there was a very big need gap for content in joyful reading content in multiple languages. So there was a child who would be speaking a certain language at home and when she would go to school or when they would go to school, there just wouldn't be any literature in the language spoken at home. And they would feel lost and the circle of failure would kind of set in pretty quickly from there. So which is why Pratham Books developed StoryWeaver as the open source platform and the journey has been very interesting. We started off in September 2015 with 800 books on the platform to now where we have 23,000 books and it's growing every day. This was the number from yesterday. I'm sure it has gone up overnight. Basically we have around 200 content and 255 languages on the platform with 40 million readership of the books that is on-platform and off-platform and about 3.6 million users who come on the platform and use it. The 255 languages that we are speaking about is the content localization, the localization of the storybooks on the platform and not the platform localization just wanted to make that clear. And again, before we dive into the whole journey of localization, we thought it was important to set context to who is it that we speak to when we localize a platform because those profiles and personas have a very big impact on the decision making. Right from language selection to down to the whole review component of the localizing process, etc. So who is our user really? So the StoryWeaver universe is basically divided into two sets. The user, the educator's librarian parents who come onto the platform to consume the content and the creator universe who are the people who contribute to the stories on the platform. So when I talk about a user, the persona that comes most to my mind is maybe an educator, librarian or a parent working closely at the grassroots level with underserved communities who are low-resourced. And they may or may not be digital natives necessarily. Just to give, just to put a face to the persona, we've shared with you the picture of Mezemir Garma. He is the founder of Rasa Bibi Library in Ethiopia, Africa. And he's right now showing the Amharic stories to his patrons from the library. Coming to the StoryWeaver creator universe, this is something that we are very, very proud of. We have a lot of translation partners on the platform and which is why the virality of language content on the platform really. So who are they? They are basically authors, translators, but also educators and librarians who feel the need of the gap of stories in Madatang languages. And they come onto the platform to create these stories so that they can take it back for their programmatic interventions with the children at the grassroots level. They may be translating offline or online. They may be engaging with the platform offline or online and then coming online to kind of publish the stories and then using them. They may or may not be bilingual with English actually. They may probably be bilingual with the dominant state language or the region language. And then the underserved language that they are serving. And definitely may or may not be, you know, we are not sure whether they're digital natives or StoryWeaver is probably one of the few platforms that they visit when they come on the internet. And mainly every communication happens via WhatsApp, etc. The photograph that you see is our favorite team from Suchna. This is a foundation, a community foundation in Virpa, West Bengal, India. They have actually translated books into Santali and Korra from Bangla books. So the source language for them is Bangla and they've actually translated over 200 books over the last two years. So over to Suman for deep diving into our journey of localization. Thanks Suman. Okay, the picture you see here is, this is what I see on my GA, the Google Analytics screen. So you can clearly see that when the visibility is huge, right? I mean, it's around 219 countries. The StoryWeaver has punched into and the content partnership is more than 50. Okay, so as of today, the numbers up. Okay, so having said this, I mean, there are while the localization, what do you say? I mean, let me set the context of localization because I have seen whenever I speak to the people, it's almost confused with the translation, only language translation. Okay, but actually localization is not just that. Okay, so localization is adapting your product or content to any market or country. Okay, keeping four things in mind. Okay, that is the linguistic part of it, cultural part of it, political and the legal differences. So it's a huge topic altogether. Okay, and translating your site language or your labels is just a part of it. However, that's the biggest part of it. I mean, when you really try to localize your product, okay, including the language, I mean, that's where all the challenges and the real Pandora bus opens. Okay, so while both localization international focus on the same goal of delivering the experience in a new language. Okay, so what is the real difference between an internationalization and the localization internationalization is the way in which you solution your product and develop it along with the ethics of SDLC cycle. Okay, so and the frameworks that you build in. Okay, and every day on which the developers work on. Ultimately, that leads into the localization. Okay, so having said this, these are the four major reasons. Okay, so while at a global level, there could be a number of reasons why people go for localization actually what is very important from story perspective is definitely being multilingual that to a scale of 259 languages as of yesterday. Okay, localization goes hand in hand, no question asked and we want to do it. Okay, that's that's a that's a need. Okay, and global reach we just talked about so too many countries we are into so definitely we want to add this into our kitty so that I mean we stand out. And when I say cultural nuances right I mean this is the actually the touching part of it. Okay, so can I for example, can I show or rather can I render story we were. Okay, so with the picture where a girl is dancing to an Afghanistan partner definitely can't because it is against the cultural fate. And so that is so when you do it when you do your localization it is not just just changing the language or taking care of currency time and all those things right I mean it is even the cultural part of it. Okay, and then coming to the platform integrations this is our strength so story we were has multiple packages offered to their partners, and those integrations definitely need this to be part of it so that I mean it really connects to them like for example I mean if it is I'll give you the example of our microsite we have a curated gateway story we were my microsite I mean that's named as learn at home you can take a look at it. I will share the references so that's localized in Hindi and English I mean Hindi currently so when you click on the story we were microsite in Hindi actually it will open story we were in Hindi, along with the curated Hindi stories so that I mean people can easily connect to. Okay, so this is one integration there are multiple more which I'll talk down in the slides. I'm now you can can you move to the next slide. Okay, so when when all of us the content team the tech team and the product management team sat together and we all said yeah the localization is the way forward and we need this to be done. Okay, what we could see is only the mountain of problems. Okay, so then we said, okay, okay, let's do then what is the way forward. So prioritization and setting up what are the real challenges and some principles for our own self so that I mean keeping our audience and end users in mind. So what we are stating is the top four reasons. Okay, so definitely the first one is continuously evolving and dynamic platform. What does it mean is I can give you a recent example. In a I think a past to past week we had three releases and take it from me at least minimum 50 levels we changed. Okay, with a new feature coming in. If that is the speed in which the application is developing how do you do that catch up game of localization. Okay, so with respect to the n number of languages. So that's where the content team the product management team and the tech team need to come together and set some principles. Okay, so that I mean we definitely cater to this big gap. Okay, so the principle that we are going ahead with in story we were is yeah there is a catch up game. Okay, at any moment in time there could be some gaps. Okay, so if you open a Hindi game the Hindi site I mean there might be some labels which are still in English which is okay, but we are going to do a periodic release only accumulating some new features and then localizing it. Okay, so I mean what I'm trying to tell you guys is I mean it's not an easy task if your product is still developing so it's only the principle that will help you to take the way forward. And now can you next one please. Yeah, this is another one. Okay, so with the shoestring budgets that we work with at this story we were right I mean there are n number of commercial solutions if I look out in the market of the shelf. Okay, so but I mean whether we want to really go for that one. Okay, or is it something you develop knowing the nitty gritties and the culture of your organization works and put in the in host frameworks. So what I'm really happy with is the process that we have come out which is really working cost effective and still the required quality has been given out. So that's another principle definitely matters right. And the third one is language support versus ROI. Yeah, this is another big topic how many languages you will you will do it when the localization they need to be a return of investment. And you need to continuously measure and take those calls. Yeah, this is not a tech person will do but the product management and the content team will take hard decisions on that. Okay, so while the tech team is just given a pipeline of saying this tomorrow you need to localize in this day after tomorrow this language. So this is another way where anybody any team who is going to work on localization has to put some principles in. Yeah, this is another one. Okay, having said that we are continuously evolving and doing multiple releases a week kind of it. So the other thing is then what is the order of localization. Okay, so where it is okay to see the gas for the end user so that the feeling and the experience is not impacted right I mean so so then what is the principle I mean it differs from different applications but we being multilingual and the kind of users I'm not told about right. So you can clearly imagine that the consumption society is more compared to the creation society. So definitely the order of localization is I mean ensure that you keep the consumption pace. Pages always up to date with the localized languages. Okay, it's okay to take the gaps on the creation or the translation side. So these are some four major areas I mean it might differ based on the application. But I'm pretty sure that these are the generic ones. Okay, anybody is going to face if the development is still on. Okay, so having said this principle said then the major challenges talked about so I'll just quickly take you people through I mean how did we still streamline the whole process. Okay, so when I talked about the localization in the beginning I said the cruxies internationalization and then only you can bring the theme of localization. So when I talk about internationalization you can go to software technology guide where it has n number of sections written on that I 18 n section right I mean I don't want to we would want to attack everything okay because it's a time and the money investment what is that you really want to be internationalized in your application. So that is another guidance that need to be set when content and product management and the tech teams need to sit together because we cannot code for everything. Okay, so from the day one. Okay, though it is supposed to be as per the three it is one time investment in my experience across the 25 years. It has been never one time investment. Okay, it's you keep developing and as feedback circle again goes back to your application development and the principles. So having said that there are four important very things anybody need to take care of when they do their software development design and the principles and architecture. So I'm not you can bring all the other three so that I'll address them together. Okay, so very, very meaningful and the very basics of it right I mean obviously you need to separate your UI elements from your source code you cannot directly tighten what if if you want to show different images. Okay, based on different languages. Okay, so if that is the case so definitely you need to bring that layer separation so that you can handle it separately. Unicode goes without saying right I mean if you are multi linguistic I mean you cannot store your Latin language in in Latin itself so it has to be Unicode I mean every storing has to happen in the Unicode format. And definitely if the whole text need to be shown in a different language, starting from the beginning the day every developer codes he or she cannot hard code anything. Every label need to be coded in such a way that I mean it comes in the resource file. Okay, at the end of the day so so you have multiple resource files based on whatever gets rendered I mean your build happens like that. Okay, and the last thing is I mean I don't know whether it will be applicable for all the applications but definitely for us because 259 languages and in number of them are right to left as well. So your CSS also need to be in such a way that I mean when you find out which language in which the sentences are rendered at any given moment in time. It has to act very neatly and intelligently to show the text accordingly. So these are the four major generic principles anybody need to at least immediately to start with. Apart from this as I said the currency date and there are so many other things right I mean which is not relevant for us. But if you ask me honestly I mean they are the easy things to solve I mean our cat or two. Okay, so having said this context right I mean I'll just quickly take you through the steps that is involved in any localization process. Yeah, the first of all I mean this whole chain will repeat per language. Okay, so first of all the scoping needs to happen whether it's an RTL language. Okay, so what is the language that gets selected and what are the nitty-gritty. Yeah, you have a internationalized code running in your source repository but take it from me definitely based on the language that you are going to select or you will be given by the content management team. You need to have the scoping done so that you upfront if you can find out some gaps and fit it in so that the whole cycle reduces. Okay, so that so that is what is I mean usually I would know I mean the technique is what is led by me we will already know what is the pipeline of the languages which will be given by Amna major is saying that something like that. So but when we pick up the language one by one and handle the nitty-gritty software some languages go just like this and some languages take a U-turn and fixing some things from the principles and then rolling out. Yeah, the next one is the trickiest thing. Okay, so it means a lot for at least the organizations like story we were like for us. I mean I have seen Amna sitting on some info messages saying that someone this is not good in India. I mean actually we rolled out one release and she immediately called me up and made me make another release just to correct one more word. One word in that. Okay, so we are as sensitive as that and because we are rendering it for children. Okay and across the across the globe. So what is really required is the proper contextualized meaning translations. Okay, so that is what it is. So the crux of it and as soon as you say that there is a language translation that's going to happen in your localization process, this step is the crux of it if you get this right. Okay, so take it from me all other phases will go very smoothly. Okay, so and it depends upon every organization how they want to set right this. I mean what we have done is there is something called story translation effective index. Okay, so I will take you through in the next slides. So I'm not measures this. Okay, so meticulously to tell me saying that French has to be in API driven. Okay, so when I roll out French using the this translation API for Hindi, she will come and say that no, it's none of the things are not working. Okay, so you I will get you the manual labels through our translators or whatsoever the means. So then we need to give her a mechanism so that the translators can work on it give us the translations for them. And then you can reload your resource files. So it depends upon the language. And that's how she measures it by going through her own. I mean we have our own decision mattresses and I'm sure when organizations do it themselves. They need they need to find out something for themselves based on the quality they're looking for. Yeah, so once you have your scoping done principle set and you know how the data is coming to you. Okay, is it API is it manual is it both of it. So once that much is done you have the whole master labels at one at your end. So then what is really required is one critical review quickly to go through it. Okay, so that I'm in the next set of processor easy and this is where the context translator sometimes it might be trans not translation right it's a translation like for example we have a word called read along in English. There is no translation API that works on it gives back the English. So that time you might need to use transliteration and in some places you might need to even transcript you might need to rewrite. Yeah. So once you get these three things right I mean, I mean on the technical side what is really left over so you have you must have an engine, which will take these inputs. Load to your resource files. Okay, and it depends upon the complexity of your application right I mean in our story we were I would like to mention that I mean our back end is on Ruby on rails and front it is on react. So you can just imagine and some values will come from DB. So you need to have three places seeded with these values, your JSON file your YML file and even the DB. Okay, so you need to find out this is where the actual tech work starts based on the application stack and the technology you will find out how do you automate from the given master translations at one point. So this is what happens and the your QA your partners your translators the content team comes out helps you out when it is loaded on any test environment to see your CSS is not breaking. Maybe some like some lines are very longer that you need to shorten them. So these are the very quick reviews that will happen before the rollout. And obviously so once the rollout is done right I mean there is a retrospection session that is definitely required I mean which I have found it very useful and effective. So you will you will go back and change your way of working or the even sometimes the application development processes as well saying that what went wrong in this rollout. So that's what I'm trying to show here that I mean there is a feedback that goes to your internationalization or you are up the way of development. Okay, so what I'm trying to show here is now that we know the whole context and the whole process how it can be attacked and with your own principles and the solutions that are built around. So it's the end to end process now. Okay, so having all the teams together. So you have the data prep we just spoke about it could be only three ways possible right I mean either you have the manual translation done or API driven or it could be the combination of it. What is really working for us is the third one manual plus API. Okay, so based on the kind of the language sensitivity context and sensitivity that we run in our story we were for some languages it is API driven. Okay, and this is what I was talking about in the previous step saying that this is what I'm nice very much worried about. Okay, if you see the due diligence that goes in right what we are trying to show you here is. Okay, so in order to conclude that for the French I can use the rather the for the Spanish I can use the French APIs. She is running the translation which an auto drafts which is coming through an API and taking the feedback from the field. And you can see that I mean it is more than 74% of the people said that auto drafts are really good. Okay, and they could make only very less changes so that the story can be published. Okay, so based on that I mean there is a principle that is been set and based on that she decides which route to be taken for the translation so that she can advise the tech team accordingly. And this is under wonderful data points so whatever the exercise that we carried out okay so we were able to help our other partner teams like MSR, Google teams and all so that which shows that okay so I mean what is the efficiency of their own translation APS right I mean in the in the real world usage. Okay, and the best thing for us is you can see that when we went for this translation APS and if they were good. Okay, actually our story publishing timeline also crunched from 45% so previously if it is taking some one hour. Okay, so there was I mean it got reduced by 45% okay that's what we are trying to show the other benefits of it. Okay, so coming back to the end to end process we have the data prep done. Okay, and content team is the main actor there and then we have the review team we just talked about it how it happens. Okay, here also measure the content team comes and sees to that I mean the nitty gritty of the transition quality has been taken care and ultimately the tech team comes and takes this and they will roll it out. So, having said this, this is this gives you the end to end process you. Okay, so this is where we are currently into and there are some improvements that I'm currently looking for or rather working on which which if somebody starting today I mean they can even keep this in mind. Okay, and start with this planning and investment itself. Okay, if further data prep I mean how beautiful it will be if you can build a localization editor itself. Okay, that means somebody should be able to open your site in a editor kind of it select a label. Okay, give the translation there itself and visualize it contextually plus the length of it or the what is the real rendering effect of it. I know I mean it's it's somewhat very advanced one but I'm, I'm sure it is not very difficult to build it. Okay, at least for the outside labels the menus and headings, we should be able to quickly build it and give it to our translator so that the tech intervention content intervention team interventions becomes lesser and it is foolproof when it is given to us. Okay, the next one in the review thing is having said that I mean knowing our partners and or which are might not might not be tech savvy. Today they're all entering the data into some central repositories. How about I mean me giving a UA for them so that I mean they can directly put into our masters. Okay, so some are very specific to story we were but you can take away the principles and see when you're starting for a location process, which one need to be really build fast. Okay, so we went in a simpler way and now we are getting into the complex improvements. And what is really possible on the rollout side is based on the text tag that we have we need to do a deployment. Okay, so whenever even a single label changes, okay, on the Ruby side, or the react side. So maybe I might need to find out a way so that I'm in today. If a one label changes 30 minutes. Okay, so because I need to do one click deployment and then then only the viewer can really see how it comes out. So that is one area we are really working on to see that I mean how can we really reduce or avoid deployment. And if I'm not changes one word it should be a real time update for her on the internal environment. So these are some things I mean with a final view of it. So this is something I just wanted to see you guys to see I mean you folks to see this. See this is We are working with a partner in Afghanistan DD library. So this is an integration at the site level. Okay, so when some some user from the DD library clicks on a story we were icon on there. So you can see that with their header and footer story we were is getting launched in their native language Farsi. Okay, through an API call which writes on the SSO integration that is single sign on. Okay, even the partner is able to tell us I mean on based on the user preference. What is the language filters that we need to set and what is the site a local that we need to even bring up to. Okay, so it could be Hindi plus Farsi filter set so that the content curation also happens. I mean this is another very strong use case that we have that why we need to go for a local localization in story.