-
-
Interface
Inspiration
Despite studying English since childhood, our team of four international students still struggles with communication. Imagine how much harder it is for someone learning a completely new language! Popular apps like Duolingo are helpful but lack the authentic, interactive experience you get from conversing with a native speaker. Plus, waiting for feedback on mistakes slows down progress. That’s why we created EVA—a revolutionary AI that provides real-time grammar corrections and immersive, human-like conversations, accelerating language learning like never before.
What it does
EVA, an AI-powered language tutor simulates real, human-like conversations to enhance language learning experience. It provides real-time feedback to learners on grammar, instantly correcting mistakes during interactions. Unlike traditional apps, EVA immerses users in a natural conversational flow, helping them improve fluency and confidence by offering the kind of interactive, engaging experience you'd get with a native speaker.
How we built it
We've implemented Simple and Effective Zero-shot Cross-Lingual Phoneme Recognition(https://arxiv.org/abs/2109.11680) to build a language independent pronunciation checker. For grammar correction, we've utilized a prompt-engineered GPT-4o model because which is uniquely suited to address the problem of grammatical and sentence structure errors due to the vast amount of training data it has been trained on. As a result, it can support multiple languages. The entire application is built upon Flask Framework, and utilizes Tailwind CSS for the frontend design.
Challenges we ran into
Challenges we faced was using the Lens Studio to create an application for the Spectacles. We faced with many bugs and features that were not working. More specifically the Text to Speech feature.
Accomplishments that we're proud of
The biggest accomplishment we're proud of was making a sudden pivot on implementing our product into a web application after trying to make the application work with the Spectacles and SnapAR.
What we learned
We learned to tackle challenge and adaptive with our approach and make the best of our situation.
What's next for EVA
We plan to include multiple languages and make the application live before the end of this semester i.e. Fall 2024. We also plan to incorporate an interactive talking face to offer better learning experience.
Log in or sign up for Devpost to join the conversation.