You probably heard of software, that makes it possible to automatically scan information through recognizing text in documents, like mobile photo scans. This software is called OCR software, or software based on artificial intelligence (AI).
Lanes & Planes creates 70% of all receipt themselves and has a 100% success rate on those receipts. For the other 30%, it was our deliberate decision to not use OCR or AI based software.
In case you wonder why we decided against this kind of text recognition software, we are happy to explain the reasons.
Variety of individual receipts
Standardized documents (such as uniform invoices) can be recognized relatively easily by OCR technology, since the underlying algorithms for so-called 'capturing' can search for recurring keywords on the scan using a simple scheme. However, the documents that our users upload as receipts via desktop version or app, do not represent standardized documents, but have various characteristics (such as taxi receipts, hospitality receipts, fuel receipts, etc.). This makes it difficult for the OCR software to recognize the text content, as it cannot be read out on the basis of known keywords/formatting. OCR software offers a fairly high recognition rate for standard documents, but is particularly prone to errors if the scan deviates from the norm.
How high do you think the recognition rate of an OCR software could be for a document like this?
The more frequently users submit receipts, the more routinely they proceed. In order to ensure that users deliberately chose to submit the stored information in this way, it is important to ensure that the user deliberately entered the data in this way and that it was not stored incorrectly due to the recognition error of the OCR software.