Automatic receipt data extraction and expense categorization present in modern receipt scanner and expense tracker apps rely on the use of OCR (reading the receipt) and AI (understanding the receipt contents), which allow the computer – or your smartphone – to automatically extract all relevant information from the scanned receipt.
In this article we will give you tips & tricks on maximising receipt legibility and accuracy when using SparkReceipt or other receipt scanner apps relying on similar technology.
What is OCR and AI?
OCR stands for Optical Character Recognition. In simple terms, OCR is like teaching a computer to read text from images. It’s a way for the computer to look at a picture of words, like a scanned receipt, and turn it into text that it can understand and use.
AI on the other hand means Artificial Intelligence. Imagine a computer program that learns from what it does, getting better and smarter over time. That’s AI for you. It’s less about following set rules like traditional computer programs, and instead it’s more about constantly adapting and improving. One of the most advanced language-based AIs today is OpenAI’s ChatGPT, which is also the AI used by SparkReceipt.
How OCR and AI Make Receipt Scanning Better
So, how do OCR and AI work together in receipt scanner apps like SparkReceipt? Great question!
Reading Your Receipts: First, OCR kicks in to ‘read’ the words on your receipt. When you scan it with the SparkReceipt app, the OCR technology turns the picture of your receipt into digitized text. This is useful in two ways: firstly, you are able to search from within the full text of the document since it is digitized. Secondly, the digitized text can be fed to a language-based AI like ChatGPT for deeper understanding of its contents.
Making Sense of It All: This is where AI comes in. It takes the text that OCR reads and figures out what it means. For example, it can tell which name is the merchant, which numbers are prices or taxes, which words are purchased items, and so on. To read more about how AI can also improve expense categorisation, click here.
Why Should You Aim for Clear and Easily Legible Receipt Scans
Even though the receipt scanner app relies on advanced OCR technologies that will do their best to improve the captured receipt’s clarity, in the end they can only do so much. In essence, the better your scans, the better the AI is able to auto-extract all necessary information from your expenses.
For best results, you can help the OCR & AI processes perform well by following these tips. And if you are in a rush and cannot always do it perfect, do not worry. The technologies are able to do pretty awesome auto-corrections on minor mistakes.
Receipt Scanning Tips
Choose a Flat Surface: To prevent unnecessary skew, it is best to choose a flat surface when possible. Although your lap or stable picture of your hand holding the receipt usually work fine too, as long as the picture is not too blurred because of camera shake.
Positioning the receipt: For best results it is recommended that you take the picture from directly above it, while keeping the receipt lines horizontally level. While the app is able to detect receipt lines and auto-rotate the picture for OCR, it can make occasional mistakes. By scanning the receipt this way, you can ensure maximum readability every time.
Optimize Lighting: Especially if you have an inexpensive phone with not-so-stellar camera, make sure the receipt is well-lit. You can activate the flashlight from the camera view if needed. If possible, avoid casting large shadows on the receipt before capturing.
Take the picture while the receipt is still new: It is no secret that receipts are known to quickly fade, crumble and lose their clarity. It is advisable to take the receipt scan immediately after receiving it. If it lingers on the bottom of your bad or purse, or is exposed to light for too long, it can quickly become illegible even to the most advanced OCR-technologies. Or even to you, making you miss tax deductions on that expense altogether.
Straighten crumbled receipts: If your receipt is already damaged, you can improve automatic scan results by straightening it a bit before capturing. This minimised the reading errors by OCR and reduces the detective guesswork needed by the AI to make sense of it.
Crop receipt to edges: When scanning the receipt, while not strictly required it is advisable to ensure the receipt is correctly cropped to edges before accepting the scan. Based on the receipt edges the receipt scanner app is able to perform perspective correction and deskew the image before reading it.
Rescan if necessary: If you get a blurry picture it is often best to retake it. It is probable that the OCR will not be able to read it properly and then the AI will have nothing to work on, basically meaning you will need to enter the expense information manually. Also, a very blurry picture might not be eligible for tax deductions if all information is not clearly visible.
Prefer digital receipts when possible: Some major chains, like Apple Stores, Starbucks and Target allow you to opt-in for digital receipts. These can be directly added to the app as PDF files, or forwarded from your email to the app. To read more why we support and embrace the digital future or receipts, read this article.
Conclusion
It’s essential to remember that the clearer your scans are, the better the OCR and AI technologies can work to auto-extract and categorize your expenses. If you follow these tips, you’ll be setting yourself and your receipt scanning app for success. So go ahead, give these techniques a try the next time you scan a receipt.