About the Classifier

Learn about how documents are classified and the factors that affect classification.

When documents are uploaded to a learning instance, the Classifier sorts the documents into document groups based on the layout, which is the order in which the fields appear on the page. If a document is uploaded to a learning instance in production and the document does not match the layout of any of the document groups, the Classifier creates a new group for that document.

The following graphic illustrates how uploaded invoices that contain specific fields are sorted into one of three groups, depending on the order in which the fields appear in the invoice.

The Classifier sorts documents based on the layout

Alternatively, you can manually define a document group to process all the documents with an identical layout. In this case, IQ Bot bypasses the Classifier and sends the document to the group you created that has the same layout as the uploaded document. This option is useful when the OCR fails to detect one or more fields in a document, causing the Classifier to unnecessarily create a new document group or send the document to the wrong group. See Create a new document group.
Note: User-created document groups cannot be deleted.
The Classifier is supported by RPA bots and the Document Classifier package. It contains actions that automate identifying and sorting the different document types. See Document Classifier package.

Resources

Watch the following videos to learn more about how the Classifier works and how to use it:

Learning instances created with out-of-box or custom domains

IQ Bot checks for a minimum number of fields that you selected to successfully classify the document. In all other cases, the document will be marked as unclassified.

For example:

  • If you create a new learning instance with one to four fields selected, IQ Bot checks for at least four fields from the total number of fields presented for that specific learning instance. Similarly, if you create a learning instance with five fields selected, IQ Bot checks for at least five fields.
  • If you create a learning instance with six or more fields selected in the learning instance, IQ Bot checks for at least six fields. For example, if you selected eight fields in an invoice domain, IQ Bot checks for at least six fields out of the total fields available.

Learning instances created with other domains

If you create a learning instance with multiple fields, IQ Bot checks for at least one field and classifies the document. If IQ Bot does not find any field, the document will be marked as unclassified.

Factors that impact classification

Factor Notes
Page layout If the page layout is different from the original document with respect to document content position, IQ Bot sends the document to a different group.
Label orientation All the labels in a document must be oriented in the same direction. If some of the labels are vertical and some horizontal, the document is sent to a different group.
Label position Labels can be positioned anywhere on the page. However, the label sequence from left to right and top to bottom must be the same as in the training documents.
Match percentage IQ Bot uses a 70% match for labels.
Fields added after the learning instance is created Fields added after a learning instance is created are excluded from the classification process.