Inspiration
Many of our team members have grown up around adults in our life who did not have access to language learning skills that would have helped them gain opportunities or face less hardship in their lives. We wanted to create something that would be able to reflexively teach any interested learner their target language, entirely accessible, personalized, and without judgment. Our inspirations were Duolingo and the Pimsleur method.
What it does
LangGPT is a language-learning app that allows anyone, regardless of language level or familiarity to develop reading comprehension, vocabulary advancement, and conversational practice in a non-native language of their choice with an AI tutor that speaks back.
How we built it
We used the streamlit Python library web aspect of project, and Python and SQL for most of the back-end. We used ChatGPT-4 for the bot’s API. A few of us speak other languages enough to be able to assess GPT’s language accuracy, so lots of personal talking to the bot testing was also involved.
Challenges we ran into
Some of the challenges we ran into was learning a web-development framework on the fly, learning to utilize and tune an open AI API model, and implementing the authentication screen. Being able to create a program that could track progress over time and making sure it was accurate while only having the time of the code being operational within the span of Shellhacks was definitely one of the more difficult parts of the data.
Accomplishments that we're proud of
The basic idea was just going to be the bot’s ability to speak. As it grew, having a consistent, useful critique and grading system plus a progress-tracking system was a very useful and important part of what we managed to accomplish this weekend. Just being able to create a functioning web app in such a short time frame when none of us had really touched streamlit before this weekend also felt very rewarding.
What we learned
We learned API calls and API integration, AI model tuning, working with SQL and a database, and how to use streamlit and its state management system. All of us were familiar with Python, but definitely learned quickly that there was much more that we could do with Python than we were initially familiar with.
What's next for LangGPT
We would like LangGPT to produce learning material for the user based on the learning style, and be able to collect more data that can aid the user’s learning process. Ideally, we would be able to include personalized and active lessons on vocabulary, syntax, and grammar with more structure for the user. We also think it’s best for scalability and user experience to migrate away from streamlit. While it was a great to get our demo working in such a short time frame, it does not provide the usability and level of design and customization that a web app like this really deserves.
Log in or sign up for Devpost to join the conversation.