I have a huge number of scanned receipts for which I need to extract some data points like receipt total, date etc. Scanning through the receipts I will typically find more than one amount, so the problem then is to guess which one is the most likely candidate
for being the receipt total.

My thought was to get a first rough estimate using some known data like Mall, Merchant and which line the current amount candidate was found to infer if the amount is likely to be the total. Like this : Given Mall X and Merchant Y, then as this amount A
is found on line L, what is the probability that it is the receipt total? I would like to express a model where Mall & Merchant are independent inputs, Line depends on those and finally total amount on all of them.

Anyone able to help? The library is so incredibly expressive that the simple beginnings are just out of reach I feel.