Decisions around classification and linking are made by AI models, which raises a logical question: why did the system make this particular choice?
In this article, we explain how Paperbox arrives at its decisions, which signals the model uses, and where in the application you can find the reasoning behind each choice.
To see more information about the choices Paperbox made to classify an email or document and link it to reference data, click the "info" ("info") icon under the "Linking" ("LINKING") and "Automation" ("Automatisering") tabs on the left-hand side:
Linking (matching with reference data)
Paperbox tries to link information from the email and its attachments to specific reference data. Particular fields and data points are used to make this match. The explanation of this link can be found across the different tabs in the explanation screen:
Summary of the link:
Technical details:
Example of an unlinked document:
Classification
Classification is the process of assigning a document type to an email or attachment. This happens in two ways:
1) Classification by type via AI classification:
Here, Paperbox determines the classification using an AI model that analyses all the information (content + data) of the documents to reach a conclusion:
2) Classification by rules:
Here, Paperbox determines the classification using hard rules, such as a match on sender domain (e.g. @expert.be):
