Checkboxes and Signatures

Overview

Checkboxes and signatures are data types for non-text fields. Hyperscience allows you to extract signature and checkbox fields from Structured and Semi-structured documents. 

The possible values for checkboxes and signatures are:

  • True - indicates that a checkbox is checked.

  • False - indicates that a checkbox is not checked.

The possible values for signatures are:

  • True - indicates that a signature is present.

  • False - indicates that a signature is not present.

Structured documents

To start extracting checkbox and signature fields from Structured documents, you need to add fields with the appropriate checkbox and signature data types to a Structured layout. To learn more about how to add fields to Structured layouts, see Creating Structured Layouts.

Identification

You can automatically identify checkboxes and signatures in Structured documents by using Structured layouts. All fields, including checkboxes and signatures, have locations on the layout. The machine uses these locations to identify all fields on a document. To learn more about Structured documents and Structured layouts, see Understanding Document Types and Determining Layout Type.

Transcription

When all fields have been identified, the machine performs transcription. During transcription, the machine automatically transcribes all fields it is confident of transcribing. You can control transcription accuracy for checkboxes and signatures, using the following flow-level settings:

  • Structured checkbox target accuracy

  • Structured checkbox threshold

  • Structured checkbox minimum legibility threshold

  • Structured signature target accuracy

  • Structured signature threshold

  • Structured signature minimum legibility threshold

To learn more about transcription settings for checkboxes and signatures in Structured documents, see Flow Settings.

Semi-structured documents

To start extracting checkbox and signature fields from Semi-structured documents, you need to add fields with the appropriate checkbox and signature data types to a Semi-structured layout. To learn more about how to add fields to Semi-structured layouts, see Creating Semi-structured Layouts.

Identification

You can identify checkboxes and signatures in Semi-structured documents by using Semi-structured layouts. To automate the identification process, you need to train Field ID models. Field ID models help you automatically identify fields, including checkboxes and signatures. To train a Field ID model, you need to upload a minimum of 400 qualified documents. To learn more, see Training a New Field Identification Model and Requirements for Training a New Model.

If...

Then...

you do not have a trained Field ID model

you need to perform manual identification.

a field is set to always go to Identification Supervision

the machine is not confident enough in predicting checkboxes’ and signatures’ locations

During manual identification, you need to create bounding boxes around the checkboxes’ and signatures’ locations. To learn more about Field Identification, see Field Identification.

Transcription

There is no transcription for signatures in Semi-structured documents. The machine uses the signature’s identified location to determine whether the field is signed or not.

Transcription for checkboxes in Semi-structured documents works the same way as transcription for checkboxes in Structured documents. Similarly, you can control transcription accuracy for checkboxes, using the following flow settings:

  • Semi-structured checkbox target accuracy

  • Semi-structured checkbox threshold

  • Semi-structured checkbox minimum legibility threshold

To learn more about transcription settings for checkboxes in Semi-structured documents, see Flow Settings.

QA tasks

QA tasks continuously improve various machine learning models and help the system to deliver more accurate results over time. You can find QA tasks under the Quality Assurance Tasks section in Tasks > Perform Tasks.

Identification QA

Checkboxes and signatures are available for Field Identification QA sampling. Field Identification QA tasks are specific to Semi-structured documents. During Field ID QA, the system shows you the predicted location for a checkbox or a signature, and you need to confirm whether the prediction is correct or not. To learn more, see Field Identification Quality Assurance.

Transcription QA

Checkboxes and signatures are also available for Transcription QA sampling. During Transcription QA, the system asks you to confirm whether a checkbox is checked or not and whether a signature is present or not. The machine uses the results from Transcription QA to improve performance over time.