-
26etMoi
Inspiration
When I saw the tagline for the TypingDNA challenge, I knew exactly what I wanted to do: A DNA genetic analysis service like 23andMe, but for texting patterns.
What it does
Like 23andMe, it finds out how similar a person's texting pattern is compared to people from other countries. The output is a percentage of similarity, categorized by the country. Since the model is trained with other people's input, the more records we have, the more accurate the model will be.
How we built it
The most important part is measuring the similarity of a text in respect to others. Once the model is able to roughly give an estimate of similarity, it's just a matter of building a simple server and client to record the input and serve the output.
Challenges we ran into
Building an effective model. The model is not optimized by any means, but it can be improved upon in the future. Also, right now the model was only trained with manually created dummy data.
Accomplishments that we're proud of
The client and server successfully takes the input, records it, and serve the predicted output to the user in a chart.
What we learned
That there are so many different ways we can calculate similarity, and there's no best way to do it. Also, I had to learn Flask to build a server for the first time.
What's next for 26etMoi
Deploy on AWS and write a Lambda code that trains and updates the model with new user input automatically at a fixed interval.
Log in or sign up for Devpost to join the conversation.