Train your learning instance

IQ Bot analyzes documents, and groups them based on similar content, structure, and layout.

Prerequisites

IQ Bot performs an initial field mapping based on existing knowledge from any pretrained document types. When the first group is created, you can start reviewing the results of the initial mapping, and train the learning instance by making corrections. There are three panels in the training window:
  • left panel displaying a list of fields
  • center panel displaying field label, value, and parameter for each selected field
  • right panel displaying the document in training

Procedure

  1. Select each field in the left panel to verify that the label of the field in the center panel are correctly identified, and the location of the value is correct.
  2. If the label location is incorrect, choose the correct one by clicking directly on the text in the training document to auto populate the text in the center panel.
  3. The Designer tries to find the appropriate value for the selected field label from the training document. However, if the displayed value is incorrect, you can either choose the correct one by clicking directly on the value in the training document, or use the Draw tool to select a bigger area for the value directly in the training document. The selected value displays in the center panel value field.
  4. In addition to individual fields, IQ Bot captures data from tables. Verify the mapping for table columns as well.
  5. Ensure each table field has a correct corresponding column name.
  6. In the Table settings, select the best column for row detection, as well as an optional end of table indicator which tells IQ Bot that anything that comes after it will not be considered as a line item. Click on the text in the right panel or type the value manually in the End of Table Indicator field, in the center panel.
  7. Enter multiple labels in the End of Table Indicator field in this format: Subtotal | Tax (USD). The | pipe symbol separates the labels.

Next steps

Once you complete training the current document, click the See Extraction Results button to review the extracted fields as well as the table line items on the left, and compare them to the document on the right. You can review other documents in this group. Since IQ Bot grouped similar documents, it can train one document and process others in the group in the same way.

The See Extraction Results menu lists all the uploaded training documents in an alpha-numeric sequence. The files are listed in old to new sequence.

As per the file names, the training documents in a batch are listed in the following order:
  • Files names starting with special characters.
  • File names starting with numbers.
  • File names starting with alphabets.