Build a Receipt Parser with GPT-3

In this tutorial, I would like to walk through how you can build a receipt parser with Tesseract.JS and GPT-3.

The idea here is we take the scanned receipt image, do an OCR with Tesseract.js and call GPT-3 to extract the receipt number, date and amount.

Receipt number: MGA480000366 
Date: 11/08/2022 
Amount: $31.36

We will see if GPT-3 is able to extract the right information.

Create NodeJS App

We will create a NodeJS app to perform an OCR on the scanned receipt image.

Install the following libraries:

npm install openai fs 

Register an account with OpenAI and get the API key.

Play around with OpenAI Playground to test and check the result.

If you are satisfied with the result, let’s continue to create the NodeJS app.