Facilitate Mandarin acquisition for English speakers, given the difficulty of learning the language.

Host our application on a remote web server, allowing open access to any users with the requisite hardware.

Demonstrate the power of commercial BCIs in allowing everyday users to develop practical life skills in an efficient and engaging manner.

Our project is based on the subsequent memory effect (SME), or the difference in brain activity while an individual studies an item they will later remember versus an item they will later forget. We plan to leverage the power of deep learning models to tap into the intricacies of the SME in EEG data, and deploy these models to make real-time predictions on whether users will learn the translations studied in the app.

Collaborating with Dr. Jacqueline Cummine, a Tier 2 Canada Research Chair in Neuroscience and Literacy, and Dr. Kuo-Chan Sun, a subject matter expert on Chinese language, we're running a research study at the University of Alberta this academic year to confirm that our methods can make accurate predictions on subsequent learning based on brain activity. The structure of this study will be as follows:

We will be running an initial study to confirm that data gathered using our hardware setup can demonstrate well-established effects seen across the cognitive neuroscience literature. This will involve running an oddball task, where a distinct waveform known as the P300 should appear in EEG data when participants see an out-of-place stimulus on-screen.

In this task, we'll have participants study different English-Mandarin translations while their brain activity is recorded by an EEG. Then, we'll develop machine learning models that can learn what subsequently remembered/forgotten translations “look like” based on brain activity.

Finally, we'll deploy the models created in the Baseline Task to make predictions in real-time while other participants study English-Mandarin translations. Then, we'll evaluate whether the predictions made by these models are informative - i.e., if re-studying words flagged by the model as “unlearned” helps participant accuracy.

In addition to offering feedback at the end of a given learning session (English to Mandarin association, and vice versa), we will offer users stats on their learning across multiple sessions via a profile dashboard as they progress in their learning, which in this case will be increasingly complex Mandarin characters. Users can get a sense of their progress at a glance and will be provided a summary of the characters that they need to study more.