ZenotiHelp

Smart Document Extraction β€” Field Definitions

ZTZenoti Docs Team Updated October 26, 2026 6 min readGuide

Smart Document Extraction lets you define the specific data points you want Zenoti to extract from uploaded documents. Each field definition tells the extraction engine what to look for, what format to expect, and how to handle missing or malformed values.

Audience: Administrators and form designers who configure document extraction field definitions for use in forms workflows.


On this page


Before you begin

  • Your organization must have Smart Document Extraction enabled. Contact your Zenoti representative if this capability is not available in your environment.
  • You need access to the form builder to define extraction fields.
  • Familiarize yourself with the data types supported for extracted fields before you begin β€” see Supported data types below.

Steps

Understand supported data types

Each field definition requires a data type that tells the extraction engine what kind of value to expect. The following data types are supported:

Data type Description Example values
Text A free-form string value (default when no type is specified) John Smith, Suite 400
Date A calendar date 2024-03-15, March 15 2024
Number A whole or decimal numeric value 42, 3.14
Currency A monetary amount $150.00, 250
Yes/No A true or false value Yes, No, true, false

Note: Text is the default data type. If you do not specify a data type when defining a field, the extraction engine treats the expected value as free-form text.

Define a target field

Each target field you define carries the following properties:

  1. Field name β€” Enter a unique identifier for the field. This name is used internally to reference the extracted value. Use a short, descriptive name with no spaces (for example, InvoiceDate or PatientDOB).

  2. Description β€” Enter a human-readable description that guides the extraction engine toward the correct value in the document. Write this as if you are explaining to a reader what to look for (for example, The date the invoice was issued, typically found in the header of the document). A precise description improves extraction accuracy.

  3. Data type β€” Select the expected data type for this field's value. See Supported data types above.

  4. Required β€” Select this option if the field must be present in the document. Required fields that are missing from an extracted result are surfaced with lower confidence scores, making it easier to identify incomplete extractions during review. Optional fields that are missing do not affect confidence scoring in the same way.

  5. Validation pattern β€” Optionally, enter a regular expression (regex) pattern to validate the extracted value's format. The extraction engine applies this pattern after extracting the value and flags results that do not match. Use this to enforce format constraints such as date formats, identifier patterns, or code structures.

    For example:

    • To require a date in YYYY-MM-DD format: ^\d{4}-\d{2}-\d{2}$
    • To require a US ZIP code: ^\d{5}(-\d{4})?$

Important: Validation patterns are applied after extraction. A value that does not match the pattern is flagged for review but is not automatically discarded. Always review flagged results before taking action on extracted data.

Add multiple target fields

You can define one or more target fields for a single document extraction configuration. Each field definition is independent β€” the extraction engine evaluates each field separately against the document content.

To add additional fields, repeat the steps in Define a target field for each data point you want to extract.

Tip: Group related fields together and use consistent naming conventions across field definitions. For example, prefix all date fields with Date_ (such as Date_Invoice, Date_Expiry) to make extraction results easier to read and map downstream.

Configure required vs. optional fields

The Required flag on a field definition directly influences how the extraction engine scores confidence for that field:

  • Required fields β€” When the extraction engine cannot find a value for a required field, the overall confidence score for that extraction result is lowered. This surfaces the result prominently during review so that staff can verify or manually supply the missing value.
  • Optional fields β€” When the extraction engine cannot find a value for an optional field, the confidence score is not reduced in the same way. The field is left blank in the extraction result.

Use the Required flag for fields that are essential to the downstream workflow. Mark fields as optional when the document may legitimately omit them.

Apply a validation pattern

Validation patterns let you enforce format constraints on extracted values. The extraction engine applies the regex pattern you specify to the extracted value for that field.

When a value does not match the pattern:

  • The field result is flagged for review.
  • The extracted value is still returned β€” it is not discarded automatically.
  • Staff reviewing the extraction result can see the flag and decide whether to accept, correct, or reject the value.

Important: Regex patterns are case-sensitive by default. If the document may contain values in mixed case (for example, YES, Yes, or yes for a Yes/No field), account for this in your pattern or use the Text data type with a permissive pattern.


Worked example

Example: A clinic extracts data from uploaded insurance authorization documents. They define three target fields:

  • Field name: AuthorizationNumber | Description: The unique authorization code assigned by the insurer, typically labeled "Auth #" or "Authorization Number" | Data type: Text | Required: Yes | Validation pattern: ^[A-Z0-9]{6,12}$

  • Field name: AuthorizationDate | Description: The date the authorization was issued or approved | Data type: Date | Required: Yes | Validation pattern: ^\d{4}-\d{2}-\d{2}$

  • Field name: ApprovedVisits | Description: The number of visits approved under this authorization | Data type: Number | Required: No | Validation pattern: (none)

Because AuthorizationNumber and AuthorizationDate are marked Required, any document where the extraction engine cannot locate these values produces a low-confidence result that is immediately surfaced for staff review. ApprovedVisits is optional β€” documents that omit this field do not trigger a confidence penalty.


Tips

  • Write descriptions as instructions, not labels. Instead of Date, write The date the service was performed, found near the top of the document. The extraction engine uses the description to locate the correct value, so specificity matters.
  • Start with fewer required fields. Mark only the fields that are truly essential as Required. Over-marking fields as required can produce a high volume of low-confidence results that slow down review workflows.
  • Test validation patterns before deploying. Use a regex testing tool to verify your pattern against representative sample values before saving the field definition. A pattern that is too strict rejects valid values; a pattern that is too permissive accepts malformed ones.
  • Use the Date data type for date fields even when formats vary. The extraction engine normalizes date values when the Date type is specified, which reduces the need for complex validation patterns to handle multiple date formats.
  • Coordinate field names with downstream consumers. If extracted values are mapped to form fields or other systems, agree on field names before configuring extraction β€” renaming fields later requires updating all downstream mappings.

Troubleshooting

Extracted values are consistently missing for a required field, producing low-confidence results.

The extraction engine may not be locating the value because the description is too vague or does not match how the document labels the data. Revise the field description to include alternative labels the document might use (for example, The patient's date of birth, which may appear as "DOB", "Date of Birth", or "Birthdate"). Then re-run extraction on a sample document to verify improvement.

A validation pattern is flagging values that appear correct.

The pattern may be more restrictive than the actual values in your documents. Review a sample of flagged values to identify the mismatch. Common causes include unexpected whitespace, mixed case, or format variations (such as MM/DD/YYYY instead of YYYY-MM-DD). Update the pattern to accommodate the actual format, or remove the pattern if format enforcement is not critical for that field.

All fields are returning low confidence scores even when values are present.

Low confidence across all fields may indicate that the document format is not well-suited for extraction, or that the field descriptions do not align with how the document presents its content. Review the descriptions for all fields and ensure they accurately reflect the document's structure and labeling conventions. If the issue persists, contact your Zenoti representative to review the extraction configuration.


Was this article helpful?
Your feedback helps us improve our documentation.