Build a payslip OCR validation program using Amazon Textract
Budget: $750 – $1,500 AUD
CREATE AN OCR PAYSLIP ANALYSIS VALIDATION SCRIPT USING AMAZON TEXTRACT
Purpose: Create a simple HTML page which has a name, company, date of payslip and amount of earnings input fields (as though a customer is entering this information themselves on an application form). Have a button which allows the customer to then upload a payslip to this page, and use Amazon Textract to OCR the payslip and confirm whether the information matches.
In summary, the key outputs:
1. Create the server on AWS to allow this all to happen.
2. Create a simple HTML page (not a pretty one) which allows entering a few fields as noted above.
3. Allow the user to upload a PDF/GIF/JPG/PNG file by clicking an "upload payslip" button
4. Immediately call a server-side script which analyses the document using Amazon Textract and returns JSON within the script.
5. Then have a separate section next in the script**** that reviews the JSON and compares it to the values input. It should create a table like this:
Item Input amount From Payslip MATCH/NOT MATCH
----------------------------------------------------------------------------------------------------------------------------------------------
Name David Jackson David Jackson MATCH
Date 30/06/2021 30/07/2021 NOT MATCH
Amount 40000.00 40000 MATCH
OVERALL: DOES NOT MATCH (this should be based on whether everything matches or not).
**** This script doesn't have to be perfect in how it analyses payslips. We will play with the script and improve it so that it works better and better with more payslips. It is more important that you build the structure and show the part of the script that basically does this. We can fix up the actual rules/script later in this section as many payslips are different.
6. Finally, it should return the output table above in (5) to the user in the HTML as soon as it is done.
In other words, the person uploads the payslip and after 30 seconds or so, they will see the table if it matches or not.
Purpose: Create a simple HTML page which has a name, company, date of payslip and amount of earnings input fields (as though a customer is entering this information themselves on an application form). Have a button which allows the customer to then upload a payslip to this page, and use Amazon Textract to OCR the payslip and confirm whether the information matches.
In summary, the key outputs:
1. Create the server on AWS to allow this all to happen.
2. Create a simple HTML page (not a pretty one) which allows entering a few fields as noted above.
3. Allow the user to upload a PDF/GIF/JPG/PNG file by clicking an "upload payslip" button
4. Immediately call a server-side script which analyses the document using Amazon Textract and returns JSON within the script.
5. Then have a separate section next in the script**** that reviews the JSON and compares it to the values input. It should create a table like this:
Item Input amount From Payslip MATCH/NOT MATCH
----------------------------------------------------------------------------------------------------------------------------------------------
Name David Jackson David Jackson MATCH
Date 30/06/2021 30/07/2021 NOT MATCH
Amount 40000.00 40000 MATCH
OVERALL: DOES NOT MATCH (this should be based on whether everything matches or not).
**** This script doesn't have to be perfect in how it analyses payslips. We will play with the script and improve it so that it works better and better with more payslips. It is more important that you build the structure and show the part of the script that basically does this. We can fix up the actual rules/script later in this section as many payslips are different.
6. Finally, it should return the output table above in (5) to the user in the HTML as soon as it is done.
In other words, the person uploads the payslip and after 30 seconds or so, they will see the table if it matches or not.