OCR and Authenticity

OCR and Authenticity information for document uploads.

Argyle offers uploaded document processing through Ocrolus' optical character recognition (OCR) service. The text within uploaded documents is scanned, and the authenticity of the document is evaluated. The results are made available to view directly in Console or retrieve via Argyle's API.

For purchasing and setup assistance, please reach out to our customer success team.

Uploading documents#

Documents can be uploaded via Argyle's API or directly by users through Argyle Link.

Document classification#

After documents are uploaded, you can optionally choose to classify the documents. This classification pre-processing step will return:

• The document type (e.g. paystubs or W-2)
• Key data fields (e.g. pay date or year)

This document information can be used to decide which documents are sent to full OCR data extraction and authenticity analysis (i.e. send only paystubs to full OCR).

Retrieving OCR data#

Console#

1. Select an individual user within the Connections section of Console.
2. When viewing Uploaded documents:
   • An OCR tag is shown when OCR data is available.
   • The Authenticity column will display the document's Authenticity score.
3. Click the {} JSON toggle to view OCR data and detailed Authenticity information.
   • The ocr_data object within the JSON will contain the OCR data.
   • The ocr_authenticity object will contain details on individual instances of tampering or other authenticity-related edits.
   Refer to Ocrolus' documentation for data field descriptions and how document authenticity is evaluated.

4. Click the View summary button to view a summary of authenticity information for the uploaded document.
   Additional OCR data including employee, employer, and income details will also be shown for paystubs.

API#

Uploaded documents can be retrieved through Argyle's /user-uploads endpoint.

The following fields contain OCR and Authenticity related information:

• ocr_data — Object. Contains OCR data.
• ocr_authenticity — Object. Contains details on individual instances of tampering or other authenticity-related edits made to the uploaded document that were detected.
• ocr_authenticity_score — Integer. Single score that captures the likelihood the uploaded document is authentic.

Authenticity scores#

Authenticity scores are determined by Ocrolus.

OCR webhooks#

Subscribe to the following webhooks to be notified of OCR updates:

• user_uploads.ocr_completed
• user_uploads.ocr_failed
• user_uploads.ocr_authenticity

OCR errors#

If an uploaded document cannot be processed by Ocrolus, an error object will be returned within the ocr_data object.

This error object will contain:

• A status field listing the Ocrolus document status. For errors, the value will either be FAILED or REJECTED.
• A rejection_reason field listing the Ocrolus rejection reason for REJECTED documents.

The two most common rejection errors are:

• DOCUMENT TYPE NOT SUPPORTED — For example, a bank statement was submitted when a paystub was expected.
• DOCUMENT IS UNREADABLE — The document was too blurry to read and could not be processed.

1"metadata": {
2    "ocr_data": {
3        "error": {
4            "pk": 80352036,
5            "md5": "e623b68dc110aad14158fdba31d40c6f",
6            "name": "form/paystubs/4e6ad54c-4232-4459-9e20-96fa721b9a38",
7            "uuid": "2cba5abd-3a2d-48a1-b786-0c6d4550d4ea",
8            "pages": 1,
9            "status": "REJECTED",
10            "document_class": "COMPLETE",
11            "image_group_pk": null,
12            "rejection_reason": "DOCUMENT TYPE NOT SUPPORTED",
13            "mixed_uploaded_doc_pk": null,
14            "rejection_reason_description": null
15        }
16    },
17    "ocr_authenticity": {},
18    "ocr_authenticity_score": null
19}

Other OCR processors#

Reach out to your customer success manager, account executive, or support@argyle.com if you are interested in using other 3rd party OCR processors for uploaded documents.