Frequently Asked Questions
Version:
How to add a new document to the one processed by Zelros?
Your document is not processed yet by Documents API? Don’t worry, it can be added!
Please provide at least 50 files by template of the wished document so that our models can start learning.
Provide also a clear specification of the fields to be labeled for LAD.
Please send your documents as you received them, don’t filter documents from the training files, otherwise bias might occur and performance will decrease. Don’t add "Specimen" documents to your files, it also introduce bias.
How to add a new field to an already existing document processed by Zelros?
We will be more than happy to add new field, and maybe someone else in the community needs this field too!
For that, we need to train our models again, so please provide us samples.
I’m using the portal and I have too many rejections: my automation rate is poor. What can I do?
I might be due to the tightness of your Business Rules. Try to loosen them a little bit ( using 'in' operator instead of 'equal') and don’t forget to consider special characters.
More details here.
I process sensitive data. How does Zelros handle sensitive data?
Data protection is at the center of our product.
API
During an API call, the document is processed and then all the data related to this document is automatically deleted once the response from the API is sent.
Not a single byte of the document is stored by Zelros, even the name of the document is deleted.
However, for quality and statistical purpose, few data is stored by Zelros such as the number of files in the document and the date&time of the API call.
You can check the data stored by Zelros by calling our Key Metrics API.
How long does it take to process a document?
For synchronous calls, with one type of file provided by document and with a size of one page, it is expected to wait from 5 to 15 seconds in LAD and add maximum 5 more seconds for RAD ( except RAD only).
For asynchronous calls, it’s proportional to the number of documents and pages in the batch sent.
Can I process all kind of documents?
We classified our documents on different criterion:
-
the number of fields required
-
handwritten or computer-typed
-
a unique template for a document or several templates of the same document
-
low-quality documents ( low resolution & high inclination of the photo) ?
Depending on the categorisation of the document, it may take more or less time to develop models in R&D.
How fraud is detected by Zelros? Is this module included in the license?
Fraud is detected by analysing of the metadata of documents and the pixel of the image (R&D).
This module is included in the pricing of the license.
Is Zelros capable of visualising correctly an upside-down document?
Yes, AI is way better at reading upside down documents than humans.
So sending an upside down document won’t impact the performance of our AI-models.
Can I use only LAD or RAD?
Yes you can, but we DO NOT recommand it.
First, because it’s the same price.
Second, because it reduces the value of automation. Using only one mode, the advisor still need to do manually the other task.
What is the difference between document_tags and document_type ?
Choose either:
-
document_type= type of the documents contained in the uploaded files -
document_tags= tag associated to the document contained in the uploaded files -
None of the above, the documents will be classified before beeing analyzed
To notice:
-
The
document_typeand thedocument_tagsavailable are listed in the user guide > Available documents, here. -
if you want to upload another batch of documents with a different
document_tags(or anotherdocument_type) you need to create another session. If you want to send a variety of documents, please don’t label the documents sent withdocument_tagsordocument_type.
Example of document_type and document_tags:
document_tags |
document_type |
|
|---|---|---|
|
|