Inspiration
When we were high school students, we had to pay a lot of money to receive quality feedback on our college essays. We felt that every student should have easy access to quality essay feedback regardless of their income. This led us to create this project.
What it does
A user inputs their essay and our model generates feedback. It explains them their strengths and weakness based on structure, word choice, theme, and the flow.
How we built it
We started by creating a Pandas DataFrame using sample essays and their respective feedback. Then we trained our model using LSTM, and Dense Layers from Tensorflow to identify positive and negative words and expressions, and whether the overall essay was good or needed improvement.
Challenges we ran into
We originally had trouble vectorizing the essays and their respective feedback. We fixed by using tokenizer, which allowed essays and feedback to be vectorized and the model to be trained.
Accomplishments that we're proud of
We are proud that our model achieved an 85% validation accuracy. It was able to identify the feedback that was given 85% of the time.
What we learned
We learned how to train a model which generates text. This was our first time creating a text to text model.
What's next for Virtual College Essay Specialist
We will try to make this model accessible on a broad range of softwares, including gmail, google docs, or into an app.
Built With
- jupyter
- natural-language-processing
- pandas
- python
- tensorflow
Log in or sign up for Devpost to join the conversation.