GPT3 with Data

Using Generative Pre-trained Transformer 3 (GPT3) to generate Python code for cleaning data, explore insights, and feature selection, then run the code.

Comment

NOTE:

** Since I ese OpenAI API secret key, so I archive the program to a zip file and save it to drive. If you cannot access it or cannot run the code because the "OpenAI key not found error" please contact me at: nguye3mt@mail.uc.edu**

Inspiration

As a Computer Science student who loves working with data and sees the important role of data in any field. I see GPT3 in particular and Natural Language Processing in general as an opportunity to help data scientists, data analysts, and businessmen to reduce the time and effort to explore, and/or clean the data. Languages are civilization. If we can train a machine that can understand our languages, there are many things that can be accomplished effortlessly.

What it does

When I start the project, I really want to make it do three things: clean data, explore data, and feature selecting. The only task I have accomplished is exploring data. The project will take an Excel (.csv) , and instruction from a user. Then generating code for that instruction. It can calculate some quantitive values, plotting, drop null values, and change the type of features. Since the limitation of time, I cannot accomplished the other two tasks.

How we built it

Using an openai GPT3 with OpenAi secret key, I create a GPT object and add some examples for how to explore data and automatically basic information like what is the table name and columns name. Then use GPT3 to generate code Using React to create a web app

Challenges we ran into

Not enough time to create a storage to store the output then GET the output put in into webapp. access the cloud service from Microsoft. App running problem

Accomplishments that we're proud of

It can calculate some quantitive values, plot, drop null values, and change the type of features. Of the limitation of time, I cannot accomplish the other two tasks.

What we learned

Natural Language Processing is amazing and a topic that everybody should learn. Moreover, it is great to explore a new modern tool that can help you in something that you never imagine

What's next for GPT3 with Data

Adding storage -> Finish Getting-Output Feature -> add more work for the model so it can clean data and do feature selecting

Built With

Share this project:

Updates