New addition to fileee - Interview with Tobias Trame. Since the beginning of the year, the fileee team has had reinforcements! Tobias, a 25-year-old mathematics graduate, supports fileee in all tasks related to optimising document analysis. The invoice amount, the address or the date on your document was not recognised by fileee? Tobias makes sure that errors occur less frequently during document analysis. In the future, you can also help us improve the analysis by sharing poorly recognised documents with us. We interviewed Tobias briefly on the topic.
What are your tasks at fileee?
Tobias: The current focus is on improving the document analysis. We are constantly optimising the recognition of individual document details, such as the date or the address. The evaluation of certain statistics is useful here. This allows us to determine how good fileee's analysis already is and where there is still a slightly greater need for improvement.
In your experience, what are the most common causes of poor analysis results?
Tobias: Our analysis is dependent on OCR recognition. If the text is not recognised correctly, the basis for a correct intelligent recognition of senders, amounts, etc. is missing. The scan quality plays an important role here. Scans with the smartphone are actually close to the quality of normal scanners, but there are a few things to bear in mind. This of course increases the source of errors. Different cameras, insufficient lighting, the wrong distance to the document can have a strong influence on the result. Documents that have been scanned with the fileeeBox, for example, are hardly affected by this, as the distance and lighting are always optimal. Other frequently occurring sources of error are, for example, unknown senders that we have not yet included in our database. Another factor that complicates the analysis is the fact that there is no standard for formats and documents are always "structured" differently. For example, it is not always clear where the addresses of the sender and recipient are located.
You are currently working on a new feature so that users can share their unrecognised documents with fileee. What exactly happens to the document and how does it help you in your work?
Tobias: First of all, of course, an error must occur, i.e. a detail of a document was not recognised or not recognised correctly. In the next step, the user then has the option of forwarding the document in which the error occurred to fileee. If a document is not fully recognised by the analysis, a request appears asking whether the user would like to send their document to fileee to help with the analysis of the documents.Let's imagine that the user sends in the incorrectly recognised document. This then ends up in a fileee test account that I can access to look at the problem. After I have identified the problem, I look for similarities and differences to other incorrectly recognised documents in order to identify the most common sources of error. If, for example, addresses are frequently recognised incorrectly or not at all, there is a specific reason for this, which I can then correct. So every single document contributes to improving the document analysis!
Before we release the analytics enhancement option to all users, we are still looking for testers who want to help us improve the new feature. Are you an iOS user and want to try out the feature before anyone else? - Then get in touch with us: support@fileee.com
