Hi I'm John McDonagh, the CTO of eKidz.eu, and I'm going to talk a bit about how artificial intelligence is contributing to a boom in language education and language development both for children and for adults. eKidz.eu is a mobile learning platform that currently has content in English, German, Spanish and Ukrainian. Coincidentally I have a 3 year old daughter, who is being raised bilingual in German and English, and also she's picking up some Spanish from her mother's side of the family.
Language problem
So, why is language development and education different from other areas of learning? Well, every child in the world quickly picks up a language without the need for lessons. The linguist Noam Chomsky was one of the first people to explore the idea of language innatism. His insight started with the observation that our native language development occurs automatically and without anything that resembles traditional teaching.
Typically a child is presented with stimuli. Say, simple sentences they hear from their parents. And once children are physically mature enough to create the necessary sounds, they start to attempt their own sentence formations. The responses from the parents are rarely explicit corrections or formal descriptions of the target grammar, but rather just general, albeit relevant, sentences from which the child can continue to attune their understanding. For continued strong language development the child will need richer stimuli from varied sources, for instance from school, from teachers and fellow students. One might reasonably ask, "If everyone naturally and automatically develops the language faculty, why would any further help be needed?"
The first reason is that even where a native language is concerned, children will have widely varying proficiencies, deficiencies and learning impediments. Children come from different backgrounds, being exposed to different stimuli and hence, will be learning different dialects. For a teacher, this makes maximizing each child's progress an impossibly complex task, since they will be working with restricted time, individualized focus and limited expertise.
When it comes to second or third languages, these obstacles to progress grow exponentially.
tech revolSo is where an exciting technological revolution comes in. For the first time in history we have the ability to start engaging children's language development both on a mass scale, whilst also with incredibly individualised guidance.
There are two reasons for this revolution. The first is the quantum leaps that have been made in natural language processing and speech recognition in the 21st century. Deep Learning Neural Networks in recent years have made massive advances in fields such as speech recognition; and we're all beginning to see the results of this in our everyday lives. So far most of the technology we commonly experience is very generalized.
But the second reason, and what really creates the synergy here - is the saturation of powerful and cheap mobile devices in peoples hands. Every child with access to a phone or tablet is a provider of the data that fuels our machine learning models. Every sentence they utter using our app helps us become better at what we are trying to do.
Imagine an 8 year old Spanish girl trying to read aloud a German text. The individualized experience should be that her competency, say with enunciating certain phonemes or with understanding certain sentence structures, should be accurately assessed by the platform, and then the most fruitful stimuli should be provided back in a constant feedback loop that maximizes the progress towards mastery of the target language.
This is precisely what eKidz aims to do. As hardware becomes widely distributed to schools, the software for home and blended learning is becoming ever more required and mandated in schools. A commensurate mandate for enhanced quality and intelligence in the educational software inevitably follows, and eKidz.eu wants to be ahead of the curve. As we build a large and diverse corpus of speech data, we also assemble a diverse constellation of algorithms, neural nets, graphs and other processors so that we can profile each child accurately and comprehensively. We constantly engage them with content that challenges them in the most productive manner possible. Of course these profiles will inform, and be informed by parents, teachers and mentors of users.
These AI and machine learning systems will be constantly expanded, updated and improved. Therefore it is necessary to have a solid and generalized underlying architecture, which can categorize, process and deliver large amounts of content, via interfaces to these current and future processors. Many auxiliary processors and tools are also essential to the success of such a solution, such as tools to clean up audio and remove background noise from rambunctious classrooms and noisy homes.
Crucially, privacy is a huge current and future concern with children’s personal data. To this extent, ekidz.eu uses secure data centers and has its own built-in encryption. Furthermore our data storage systems are organized in such a way as to allow for federalized data centers for countries or school districts, who request isolation of their own data, as well as specific geographic location of their dedicated data centers.
In summary, the vision for ekidz.eu is a personalized engagement with students from any demographic or background, and targeting a growing range of languages. Not only will ekidz.eu continue an exciting engagement for the student, but we will also keep an ongoing conversation and information exchange with teachers and parents.