Avalinguo: English fluency for Latin Americans with VR

English has become a global language, particularly in professional settings. But In Latin America, the proficiency level for spoken English is too low, and this limits our professional development. Even popular apps like Duolingo fail to promote verbal fluency.

We propose a different kind of app for learning to speak English: “Avalinguo”. It is the first smartphone app for learning to talk in English where users appear represented as avatars in a virtual room, using the cellphone.

Avalinguo could have a very low cost for the student, and it could be even free when a native English speaker volunteers to tutor a lesson or exchanges it for a Spanish lesson. Avalinguo does not require a certified teacher, as there will be guides for leading each session, with helping videos to initiate the conversation.

Avalinguo could put English fluency to the reach of millions of Latin Americans.

The problem we are solving is that in Latin America (population of more than 600 million) the proficiency level for spoken English is too low, and this limits our professional development: the English Proficiency Level has been evaluated by EF as low (50.34).

But improving the speaking fluency needs real talking practice, which is not provided by most language-learning apps. We are proposing a software that will allow Latin American people who have access to a smartphone (currently 260 million) to practice talking in English in virtual rooms with the help of native English speakers. For most Latin Americans it is very difficult to find English native partners to help them improve English talking fluency, so our solution could be extremely valuable.

Further, the use of virtual rooms reduces the need for a good Internet connection, which is a common situation in Latin America, because we don’t use video, but instead each participant sees the other ones as avatars together in the virtual room. Avoiding the use of video also promotes privacy and makes the user more comfortable when speaking to strangers –one cultural trait of Latin Americans is that we are very self-conscious, so we feel embarrassed easily.

The population we intend to serve is composed of the 260 million Latin Americans who have access to a smartphone, and who need to improve their English fluency level for professional development. We ourselves the founders have struggled a lot with English fluency, and we have validated with language experts that the talking aspect is the most underserved one, as well as the most difficult to take care of.

Our solution (virtual rooms where users appear as if they were in the same place and talk in real-time with the help and guidance of a native English speaker) provide a simple yet powerful way to practice talking on a regular basis, even several times per week.

The privacy brought by the avatars instead of the video could also help (in the view of expert polyglots) to make users feel comfortable with making mistakes while talking, especially with shy and self-conscious people, and this privacy also allows the users to have the sessions anywhere, not only in a desk, properly lit place, nobody else around, and other restrictions of practice with video.

Avalinguo is the first smartphone app for learning to talk in foreign languages where users appear represented as avatars in a virtual room, using just the cellphone. Each user can be anywhere in the world, but the ones connected to one virtual room appear on the screen of the smartphone as if they were meeting around a table. Please watch the short demo video in other question.

The Avalinguo platform will support a big quantity of virtual rooms, but each virtual room has capacity for only 5 participants (including the facilitator, who is a native English speaker). According to language experts, it is best to have conversations in small groups.

When a user who is connected to one virtual room talks, everybody in the room hears her/his voice, which is transmitted in real-time over the internet (VOIP), and everyone sees that the talking avatar is moving the hands, head, etc., which makes easy to identify who is talking. The movement of the avatar is produced artificially when detecting a voice in the microphone.

Our prototypes are just software programmed in the Unity videogame platform, but the user side can run in a standard Android or iOS smartphone, there is no need for a game console. We run the server part of our software on cloud servers (like Amazon Web Services), which have very low cost and are able to serve many users concurrently.

We have added to our prototype support for streaming a video (any of the available ones on YouTube, Vimeo, etc.) simultaneously to all the participants of a given virtual room, in order to initiate the conversation; for instance, the facilitator asks each participant what was the situation about, etc.

The app will present to the language student a list of available rooms, filtered by level and ordered by a number of factors, like the interests of that student (as declared in the profile, like sports, books, movies, food, etc.), the number of participants already in the room, etc. The rooms list would be presented like Google search results, where the top results are the best recommendations. 

In the future we will introduce advanced IA capabilities, like assessing pronunciation and fluency using Machine Learning, suggesting discussion topics, sharing lesson guides among the facilitators, and many others.

Avalinguo’s innovations are not for the sake of innovation. The speaking aspect of learning English is the most underserved one, and we wanted to make it interactive and engaging, private, practical and flexible. We proposed specific ways of applying Virtual Reality to get engagement but avoiding awkward exposure of the user and her clothes and place, as in video conferencing, and put an avatar in a virtual room instead. We also found ways to use Artificial Intelligence for improving convenience and engagement, like automatically matching each user to others in virtual rooms so that they share interests and also their participation styles are compatible (for instance, don’t form a group with just quiet people). Even the level of fluency and pronunciation is assessed by an Artificial Intelligence system, which has already been the object of scientific publications.

Avalinguo innovation is not only in the product but in the business aspect as well: we developed a two-sided market model where there are incentives for each side’s participation. We improve the economic aspect of the teaching part, compared to 1-to-1 classes, because our small groups have 4 students instead of one. We also have the option of leading a session of a Spanish-learning group and get credit for taking lessons in English groups. Quality of session facilitators is assured by a rating system (typical of any share-economy service like Uber), instead of traditional schools’ certifications.

Anyone with access to the internet can find a wealth of resources for learning a language, but several studies have found that a student learning a language in isolation ends up dropping it despite an initial high motivation. Avalinguo will be the virtual place where English students meet daily, even for a few minutes, so they renew their motivation and continue learning. The impact of talking to people for countering isolation has been proven in related studies.

In today’s cities, time is a scarce resource, so for busy professionals it is very difficult to find the time to commute to a school or the money to pay private tutors several times a week, but expert polyglots have found that study frequency is critical: a lesson a week or so just won’t cut it. This is why in Latin America most English learners get stuck at a low level, beyond which it seems impossible to advance. Avalinguo provides a convenient way of getting in daily touch with the target language, and particularly with the active speaking part of it, which is the one with less resources available even on the Internet. To learn to speak in English, you have to actually speak in English, according to experts.

We are currently serving less than 100 people because we are still testing and adjusting the system.

In one year we will serve a few thousand people, mostly in Mexico, but at that point we’ll be in a growth stage, as well as the development of some complementary functionalities of the system, so from then on, growth could be very fast.

In five years we will serve a six-figures number of users, in several countries. At that point Avalinguo will make a significant impact in the English level of the Latin-American population, which will start to reflect in the region relevant statistics.

Our goals for the next year will be to field-test the basic assumptions of our solution with real users using the product and actually learning English. It would be enough for this to have an engaged community composed of a few thousand Mexican users, as well as a small international group of people learning Spanish. There will be different ways Mexican users can participate: to learn for free, by exchanging Spanish lessons for English lessons, or to pay very little, or even getting an income by facilitating Spanish lessons.

In the following four years, we would focus on growth, mainly inside Latin America (additionally to the international Spanish students). The expansion to a user base in the order of the hundreds of thousands will be possible because of the scalability properties of our model. As the number of students grows, the number of facilitators will catch up, mainly because facilitators are not employees of Avalinguo, but freelance native speakers of the target language. We won’t even need to get computational equipment, as everything will run on cloud services. At that point, we will significantly contribute to closing the work skills gap, which in Latin America is among the largest of the world, by giving language skills to a significant part of the population. Further, the lesson facilitators could later become language teachers, so Avalinguo would be a very formative experience also for facilitators, besides giving them English instruction or additional income (or both).

We can’t be certain of all the barriers that we’ll encounter in our way to make a dent in reducing Latin America’s work skills gap, and we’ll figure out how to solve them during our journey. To begin with, one big challenge is that we’ll have to juggle many moving parts: the software development and addition of advanced AI features (like optimal group formation, automatic fluency assessment, automated discussion triggers, mini-games), but also the sustainability of the project (getting some income and perhaps investment to finance growth and features) and yet managing a two-sided market (students on one side, facilitators on the other). All of this has to grow in a coherent way, and this is a huge challenge.

The technical challenges are manageable (we have already developed most of the needed technology), but for sure we’ll meet legal, cultural and market issues. The most dangerous gaps are the things that we don’t know that we don’t know, but for sure we’ll meet at some point.

We are going to refine the product gradually, at the same time as we also grow the business aspect. This is why we haven’t fully developed all the software bells and whistles like the AI functionalities: we don’t want to invest too much on the product before we can test each of the features directly against the market, by doing experiments.  Nevertheless, the AI features are going to be very important as differentiators from the unavoidable copycats that appear once we start getting some success. 

For legal barriers, we’ll have the support of external legal advisors.

Finally, to address the cultural differences between Latin American countries, we’ll join forces with local actors (as the U. de Los Andes in Colombia) and adapt to the local culture gradually, instead of conquering the market by brute force. In the end, all Latin Americans have the same goal, which is to reduce the work skill gap, which is hindering the prosperity of our countries.

We are currently testing and refining the Avalinguo minimal prototype, which provides only two functions: to join a virtual room for 4 students and one facilitator, inside which the participants can talk and see and hear the other participants, and also to watch public short videos from YouTube, as a way of starting a conversation, checking listening comprehension, etc. We are developing guides for conducting the sessions, including what to ask the students to do or see at specified times (in the future, session guides will be shared among the facilitators community).

Our tests are being conducted for free with groups of Mexican students learning English, and with a professional English teacher who plays the role of a facilitator (though in principle the facilitator in Avalinguo doesn’t need to be a language teacher). From these closed beta tests we have detected bugs and limitations that need to be addressed before going public. 

We insist that we don’t want a perfect product at this stage, but of course we need a barely usable product. For instance, in one of the test sessions we had a nasty sound echo problem that made it impossible to understand each other. We think that if users find something like this at the first use, they would hardly try it again.

After we complete the basic refinement, we’ll try some distribution channels, like affiliate marketing for Spanish talking practice. We think it’s much easier to get Spanish facilitators here in Mexico than English native ones.

N/A

One full time:

Dr. Ramon Brena, LinkedIn: https://www.linkedin.com/in/ramon-f-brena-a8a4112/

3 part-time partners, who are the founders of Ensitech (ensitech.com):

Dr. Eduardo Ramirez, LinkedIn: https://www.linkedin.com/in/ehramirez/
Ing. Luis Canché, LinkedIn: https://www.linkedin.com/in/lcanche/
Ing. Jorge Ayala), LinkedIn: https://www.linkedin.com/in/jorgealbertoayalarascon/

One Unity programmer (Pedro Cárdenas), LinkedIn: https://www.linkedin.com/in/pedro-cardenas-benavides-38617a32/

First of all, we understand the problem to solve because we have struggled ourselves to get reasonable proficiency levels in English, and particularly at the speaking skills. Years ago, despite having taken English classes at school, we were just unable to have fluent professional communication with native English speakers. Now we are fluent, but we know the struggle of achieving professional English fluency.

Besides, the founders of Avalinguo have each one technical or management strengths that are going to be applied at Avalinguo: Two of us (Ramon, Eduardo) have a Ph.D. and research contributions in Artificial Intelligence (remember the AI features of Avalinguo). Three of us (Eduardo, Luis, Jorge) partnered 15 years ago to found a successful Software development enterprise (Ensitech), so they are experienced both in business development and in software technology –which is critical as Avalinguo is basically a software product. Further, one of us (Jorge) is an expert in digital marketing, which will be important in the growth phase of Avalinguo.

We have an ongoing collaboration with the Tecnologico de Monterrey. For instance, the AI-based methods for automatically evaluating the speaking level of an English learner have been refined in the context of Master Thesis by graduate students (Evelyn Z. and Aristh V.) and Summer projects (Alan P.). We have also coordinated Avalinguo-related activities in the INCmty festival, organized by the Tec. Our main language-learning advisor (Colette Pawlonka) is a professor at the Tec. We were selected by the Tec to participate in a two-week Bootcamp at Madrid in June 2019, organized by the Madrid Innovation-Driven Ecosystem (MIDE). We would like very much that the Tec would be among the first organizations to benefit from the use of Avalinguo to raise the English fluency of its own students and professors, and we make this extensive to the U. de los Andes.

Our business model is based on subscriptions over a two-sided market, with several additional modalities. The two sides of the market are the students and the facilitators. Normally, the students pay through their subscription and the facilitators get paid for their services, but it is also possible that a person is a facilitator of Spanish sessions and also a student in English sessions. In order to consider this and many other modalities, we are introducing a virtual currency, the “avas”, of value only inside the Avalinguo system, and when you subscribe as a student you get an amount of avas from which each session’s cost is deducted; similarly, when someone facilitates a session, gets paid in avas that can be converted to actual money if you have a positive balance. There will be many promotions, like if you refer another subscriber, you’ll get some avas, and so on.

Our distribution channels will include affiliate partners: existing YouTube channels or websites with language-learning content will receive a commision when they enroll a subscriber through a landing page. Our costs are related mostly to development, but operational ones are minimized because we won’t have our own computers or even offices (the Avalinguo software runs in the cloud, and all of our employees work remotely or are freelancers), so the operation costs are only increased when we have many clients, which is a good situation to have.

For increasing our social impact, we are going to provide special conditions for social development programs.

We are not going to rely financially on donations or grants, because we’ll have many income streams, the main one being composed by paid subscriptions, bought either by individuals (at a cost similar to Netflix) or by enterprises, associations or governments at negotiated prices. We’ll get individual customers with the help of affiliate partners and also by digital (inbound) marketing, and corporate customers by one-by-one sales (outbound marketing). We have calculated financial projections to 5 years, time that we consider enough to have established the service in several countries, have enrolled more than 100,000 users, and have yearly revenue of 7 figures (dollars). 

Other income streams will include: the rent of virtual rooms to private language teachers or language schools (a SaaS variation in which we don’t provide the facilitators, but only the digital service); or to put paid advertisements in the walls of virtual rooms; or to offer paid training for new facilitators; or to offer differentiated price tiers depending on perks like branded decoration of virtual rooms, minimum facilitator evaluation, rooms exclusive for members of an organization, etc. Special conditions will be given to social programs, as in these cases the goal isn’t to get profit but to make a social impact as large as possible.

The TPrize Challenge is a great opportunity to put to the test our proposal, from the points of view of its potential impact for our community, its financial sustainability, and its overall meaning for closing the skills gap in Latin America, today the widest in the world. 

We take very seriously the TPrize motto “Skills are the global currency of the 21st century”, and we think that language skills are critical in a globalized world. Not being fluent in English is a very limiting factor for successfully doing trading negotiations and many other real situations of great impact on the development of the Latin American region. We are aware that our solution is not oriented to the poorest or most isolated communities (those with no internet connectivity), but we think we should take a broader perspective and think about the skills development for a very sizeable part of the population (260 million people in Latin America with smartphone connectivity), and think about the way the region as a whole would close the skills gap and foster its economic and cultural development.

In case we are selected among the finalists, beyond an eventual financial help, the endorsement of the prestigious universities behind this event will give us huge credibility that will help us to get valuable partners along the way. Currently, we have no partners in Colombia, and the support of the U. de Los Andes would be invaluable to make an impact there.

We would like to launch Avalinguo pilot programs with universities, in particular the Tecnologico de Monterrey and the U. de Los Andes, so that we can measure the use and adequacy of our solution at scale. We think that their respective Languages departments could consider the use of Avalinguo for language students as a complement of their regular activities, and give useful feedback about the resulting experiences, which ideally would be registered and measured. We think that the speaking practice brought by Avalinguo could multiply the usefulness of the regular language classes taught at these institutions.

We'd also like to try our proposal with the involvement of grassroots communities or collectives, additionally to our commercial partners, like the affiliate programs, that we think we can develop on our own.

We will develop multiple commercial partnerships with well-established internet-based language schools, because most of them offer only online content, not speaking practice, which is exactly what we provide, so combining our offers through affiliate program partnerships is very natural.