The ETP Transformer

The Project

The ETP Transformer is an application with a web interface that can write Python code based on the user's English commands. The entire application was built on the integration of the Transformer neural network model. With just a small sample of training data, the EP Transformer had achieved a better and more efficient performance compared to traditional NLP models.

View this project on YouTube

Team Comments

I choose to make this project because I want there to be an educational and simple AI that can provide people with programming examples they are unsure about. With just a simple command input, people can access code examples easily without wasting their time on searching on the internet.

I found it really difficult to create an AI that can performed well with small amount of trainings. The traditional NPL models and CCNs are too burdensome to achieve my goal. So after some researching, I decided to use the newest Transformer model published by Google, which work nicely well.

If I had chance, I would definitely feed more training datas to the ETP Trasnformer, because the more data the better attentions Transformer model would have and thus performed better. Another thing I would do will be to improve the user interface of the application.