Inspiration
When I was teaching English in South Korea as a volunteer, I noticed that the students would excel in their listening, writing, and reading skills but performed far worse when it came to speaking. It turns out that they are unable to practice speaking English because of the lack of native English speakers in their vicinity.
What it does
AITA is an English learning platform that uses the power of AI to enable students to practice speaking English and practice
How we built it
On the frontend, we use Nextjs with Pandacss. On the backend, we used Java and Postgresql and deployed it using AWS.
For the conversational bot, we use Gemini API for the:
- Conversation response
- Explanation of mistakes
- Grammar correction
Challenges we ran into
We learned that students have a harder time engaging with AI from a conversational perspective, versus actual people because they sometimes find talking with AI boring. This is because the speech-to-text, despite recent advances, is boring compared to the fluctuating emotions of a person. In order to circumvent this, we added a temporary gamification feature that incentivizes the user to more frequently engage with the bot.
Accomplishments that we're proud of
We are proud of the value that we were able to provide to students learning English. For example, one of the users was able to get a score of "AL" in the OPIc exam, which is the highest grade for that exam.
What we learned
We learned that Gemini performs better (and faster) for more creative tasks compared to other models. When we first used OpenAI GPT4, the streaming output was much slower compared to Gemini, and the content of the output was less engaging. Once we swapped out OpenAI for Gemini, creativity of the conversation increases as well as the speed of the response.
What's next for Aita
We will add a feature that describes the setting and the facial feature of the bot, to add an imagination element (similar to that of reading a book), to increase conversation engagement.
Built With
- gemini
- java
- nextjs
- pandacss
- postgresql
Log in or sign up for Devpost to join the conversation.