Create OCR centered proof of concept

General information for the business: validating a business idea
Kind of development: New program from scratch
Description of requirements/functionality: Ultimately I want to create an online app that will take bills, and other unstructured scanned paperwork / documents and populate a database with the data.

For the proof of concept, I want to show that we can take documents and dump relevant content into a table displayed on a website in real time.

The scope could be limited to some predetermined templates

I see two ways of doing this,

1. OCR with vectorisation, i..e the page is split into a grid and we tell the software what each element in the grid aligns with.
2. OCR with text matching, i.e. we look for keywords on a page and associated nearby content with the content we are looking for

3. any ideas you can come up with
CMS and Admin requirements: Ability to add a file
Purge the database after each file upload
display the file content in a table
Specific technologies required: does not matter for this but should be able to run on a unix/linux web host
OS requirements: Linux
Extra notes: attached is an example water bill which would need to be converted. output in a table could be:

1. name of supplier
2. payment period
3. payment amount
4. payment type
5. payment due date
etc