Project Beluga is an open source application that allows one to compare data between a purchase order and a corresponding invoice and identify discrepancies in the data between the two documents. To get started, please provide data for the invoice and purchase order.
See a sample invoice and sample purchase order data file to get an idea of how data should be formatted.
Once the data is in place, run a comparison to have ChatGPT perform a "fuzzy matching" between items from the two sources based on their descriptions. Even if the descriptions for an item aren't exactly the same between the two documents, ChatGPT can perform an educated guess and suggest matches for items that it thinks are associated with each other.
ChatGPT has also demonstrated the ability to match text descriptions across two languages. Try uploading the sample Korean foodstuffs invoice (English) and corresponding sample Korean language purchase order data file (Korean) for a demonstration. Note that you may need to adjust the data extracted from the invoice before performing a comparison. Also note that the invoice document scanning feature currently only supports English and other European languages.
Note that any unsaved data will be lost upon closing the browser tab or refreshing the page. Data is saved in your browser's local storage.
Upload a CSV containing purchase order data or start adding rows manually. If you upload a CSV, you will be able to make changes to the extracted data after uploading.
CSV should include data in the same order as the table (Product Code, Description, Quantity, Unit Price, Total Amount). Please do not include a header row in the CSV.
To upload a file, drag it here or click here to select from your computer. File must be a CSV.
Previously Saved Data
Add invoice source data by uploading a document for data extraction by Google DocumentAI. You will be able to make changes to the extracted data.
Drag invoice file here or click here to select from your computer. Supported file types: JPEG, JPG, PNG, BMP, PDF, TIFF, TIF, GIF. Max 15 pages.
Previously Saved Data