1

Can anyone recommend a program that will allow me to scan supermarket receipts and break up the food bill into meat, dairy, vegetables, fruit, alcohol etc so that I can see where all the money on my food bill is actually going? I want the program to be able to use the categories I create on the next scan, without me having to do this manually every time I scan a receipt.

The problem with every software I have looked at is that it allocates your supermarket to Groceries. None of them seem to provide a solution to make it easy to analyse your monthly food bill by breaking it up into Vegetables, Fruits, Dairy etc.

Given that food is a major monthly expense, it makes sense to see if you can reduce your expenditure on this, but to do this, one needs to see how much you are spending on diary, meat, ready made meals etc.

Can anyone recommend a solution?

4 Spice ups
2

You could look at using Quickbooks Online OCR software (Klippa):

3

The problem with this request and the reason nobody has made software of this nature is that no two grocery stores will call a can of corn, “corn”, simply too much variation in names for the same product.

That said I will repost the answer from ChatGPT which states in more detail the complexity of your request.

To automate the procedure of scanning a grocery store receipt and sorting the items by category, you would need to develop a custom software solution. Here’s a high-level overview of the steps involved in automating this process:

Receipt image acquisition: Build a component that can capture or receive the digital copy of the receipt. This could involve integrating with the grocery store’s mobile app or setting up an email-based receipt processing system.

Optical Character Recognition (OCR): Utilize an OCR library or API to extract text from the receipt image. Popular OCR tools include Tesseract, Google Cloud Vision API, or Amazon Textract. These tools convert the scanned or photographed image into machine-readable text.

Text cleaning and preprocessing: Develop algorithms or utilize text processing techniques to clean and preprocess the extracted text. This may involve removing noise, correcting errors, normalizing text, and ensuring proper separation of items.

Item categorization: Implement a categorization algorithm that can assign items to specific categories based on their names or descriptions. You can use keyword matching, natural language processing (NLP), or machine learning techniques to automate this process. Training a custom machine learning model using labeled data can improve accuracy over time.

Category-based sorting: Once the items are categorized, sort them based on the assigned categories. You can use sorting algorithms or database queries to group the items accordingly.

Output generation: Generate a formatted output, such as a table or list, containing the sorted items categorized by the respective categories. This output can be saved in a suitable file format (e.g., CSV, Excel) or displayed in a user-friendly interface.

Error handling and user feedback: Implement error handling mechanisms to handle cases where the OCR or categorization process fails or produces inaccurate results. Provide a user interface or feedback mechanism to allow users to review and make adjustments if necessary.

It’s important to note that automating this process requires a solid understanding of software development, OCR techniques, text processing, and possibly machine learning. You may need to leverage existing libraries, APIs, or even custom models to achieve accurate results. Additionally, continuous testing and refinement of the automated system are crucial to ensure its reliability and accuracy.

1 Spice up
4

Oh, did you ask Chatgpt as well. I did too. It gave me some suggestions, but ultimately I encountered the same wall in that none of them really offer to the level that I am looking for.

Don.W wrote:

The problem with this request and the reason nobody has made software of this nature is that no two grocery stores will call a can of corn, “corn”, simply too much variation in names for the same product.

Yes I do appreciate that point and I am not expecting the program to be able to handle that. What is doable however, is to enable people to add corn from one supermarket to tinned food and add can of corn from another supermarket to the same category and then on subsequent scans it already has that information. Most people buy from the same supermarkets and tend to buy the same staples. So it really is just a question of spending a couple of hours adding all the items to categories manually the first time and then on subsequent scans to do this automatically. Effectively you just need to train the system, which can and has been done for years. So it’s not impossible, in principle if one accepts that you will need to spend a lot of time in the beginning setting it up.
What Chatgpt is suggesting is not commercially viable given the number of supermarkets around the world, the different names used for the same products etc. Machine Learning is not the route I would go for a product that most people would only pay a couple of pounds a month for. For businesses would make sense though.

Klippa actually looks promising. I will investigate that in more detail to see if it actually can do what I want to achieve. It would depend on how the machine learning works and whether categories or results can be amended. It basically looks like it does what Chatgpt suggested.

5

I have sent off an email to Klippa and Zoho on Friday regarding the particular product that I am interested in. They appear to offer the functionality that I am looking for. I am just awaiting their response.