L2Z

Inspiration

Inspired by a group member who learned English as a second language but never had an app to help him in English the way he want. He realized he needed a system to understand English better but only in the relevant parts of his daily life.

What it does

Our app, L2Z, helps L2 English learners grow their mastery of the language in whichever area they most need to learn. We found that L2 learners don't need to have perfect grammar and vocabulary but have specific use cases in their daily lives that they need to use English. We help them reach their goals in less time by generating context-based vocabulary lessons using natural language processing. Based on a provided context, we are able to generate a list of vocabulary that is heavily used within that context that they may not be familiar with. We then form these words into sentences along with words they already know in the context they provided to help their learning.

How we built it

We built L2Z using a variety of language models to perform natural language processing with words from the English Language. This required a lot of data transformation and cleaning as well as prompt engineering to get the output data that we were looking for. We also built a front-end using the flutter framework so that we could be cross platform.

Challenges we ran into

We ran into two main challenges:

1. Misformatted data: We found a lot of data that we wanted to use for our project and eventually did but due to difficult to translate format that kept us away from even more data that could have enriched our project.
2. System Design: This was a difficult but interesting part of our design. Since we were using an api for some of our natural language processing we had to make sure that our design was efficient in how many times we invoked the api and specifically how many tokens we requested from it as they can quickly scale and start costing actual money. ## Accomplishments that we're proud of We are proud of getting all of models working correctly to produce the contextualized sentences with new words. We are also proud of building a front end that is interactive and elevates the user experience while learning. ## What we learned We learned how to design a system using natural language processing that can scale. We learned how to ideate fast and think deeply within the constraints of a competition. ## What's next for L2Z The next step for is to connect the frontend to the backend instead of manually moving the data we generated from natural language processing. We also want to continue to build out our frontend so user can have a more complete language learning experience while using the app.

Built With

• flutter
• gpt-3
• python
• spacy