Inspiration
Learning a language is hard, especially if you have no one to talk to in the language or are too shy to. To address this, we wanted to create a way for people to receive feedback on their language skills while they speak in realistic scenarios.
What it does
Language learners often lack real conversation practice, low-stakes environments, immediate grammar correction, and exposure to real-world scenarios without travel. LingoLearn fills that gap with an interactive AI roleplay system. Our goal was to create a system where users could pick a language, a difficulty, and a scenario (like ordering food, sitting in a taxi, giving a speech, etc.), and then start the conversation. The app would respond while in character, while providing feedback on grammar as an external observer. At the end you would receive a transcript of the conversation and the feedback you received. So far, LingoLearn allows you to have a conversation with the app and receive feedback, but users must type out their responses. The transcript is displayed as the conversation progresses since the conversation is typed out.
How we built it
We built LingoLearn using HTML, CSS, JavaScript, and Gemini API. The UI displays all the content, while Gemini API was used to process the user's side of the conversation and provide replies and feedback.
Challenges we ran into
A challenge we ran into was converting the API's response into a JSON, since it returned as text. However this was a quick fix once we realized what was going wrong. We also had some trouble at first coming up with an intuitive UI and actually implementing that UI. With so many things going on in the app, it was hard to prioritize what we should focus on.
Accomplishments that we're proud of
We are proud that we were able to still get a working conversation going and receive feedback as they talk to the app.
What we learned
We learned how to use LLM APIs, which we didn't really have a lot of experience with before but is a really useful skill.
What's next for LingoLearn: Conversational Language Coach
We would like to finish implementing the speech feature so that the conversation can be spoken and not just typed, for a more realistic feel. We would also like to add the rest of the scenario options and add more languages, as we had originally planned.
Log in or sign up for Devpost to join the conversation.