Rewrite

• 

  Training Data

• 

• 

Inspiration

We were inspired by how common negative self-talk is, especially for young people dealing with stress, school, and competition. A lot of us have had moments where we think things like “I’m not good enough” or “I have no one.” We wanted to create something that could help people pause and rethink those thoughts through something simple, accessible, and encouraging.

What it does

Rewrite is a program that takes a user’s negative thought and returns a more positive or constructive version of it. Users just type in how they’re feeling, and the app reframes their thought using a softer, more supportive tone, similar to how a friend or therapist might respond.

How we built it

We trained a machine learning model using large datasets of negative thoughts paired with their positive counterparts. We used Google Colab for training, which allowed us to access the GPU resources we needed.

Challenges we ran into

A major challenge was finding a platform with enough processing power to handle training. We originally started in Visual Studio Code, but the lack of GPU support made the process too slow. This led us to pivot to Google Colab, which gave us better access to high-performance resources. On top of that, working under tight time constraints meant we had to simplify some features and focus on the core functionality.

Accomplishments that we're proud of

We’re proud that we were able to turn an idea into a working, polished web app in such a short amount of time. We created a tool that can genuinely help someone feel better, and we did it while improving our skills every step of the way.

What we learned

We learned how to train and work with machine learning models, and how to process large datasets for a specific goal. We also learned about the challenges of managing computing resources and collaborating under pressure.

What's next for Rewrite

Going forward, we’d like to connect the app to an API so that it can generate personalized suggestions for further steps based on the user’s input. This could include motivational messages, coping strategies, or online resources tailored to the specific thought. We also hope to improve the model and expand the dataset to handle even more types of input.

Built With

• collab
• llm
• python