Labeling Behavior

Controls how closely a document label must match a defined label in the Label Set for extraction to succeed.

• Value is a decimal between 0.01 and 1.0 (where 1.0 = exact match).
• Lower values allow for more flexible (fuzzy) matching, accommodating typos or OCR errors.
• Higher values require stricter matches, reducing the chance of false positives.
• Typical values: 0.8–0.95. Adjust as needed for your document set.

Tip: If labels are highly consistent, use a higher value. For variable or noisy documents, use a lower value.

Controls the similarity threshold for matching table or section headers in Label Sets.

• Value is a decimal between 0.01 and 1.0.
• Used by extractors such as Tabular Layout and Row Match to identify header rows.
• Lower values allow for more flexible header matching, which can help with inconsistent or OCR-impaired headers.
• Higher values require headers to match more exactly.

Tip: Tune this value if table or section headers are not being recognized as expected.

The 'Weightings' property allows you to fine-tune how the fuzzy matching engine evaluates differences between document text and Label Sets.

• Adjust the cost of character swaps, insertions, and deletions to better handle common OCR or typographical errors.
• Use this to improve extraction accuracy for documents with frequent or predictable misreads (e.g., 'O' vs. '0', 'S' vs. '5').
• Applies to both label and header matching when fuzzy matching is enabled.

Example:
If your documents often confuse the letter 'O' and the number '0', you can lower the swap cost for this pair, making the extractor more tolerant of this error.

For more details on configuring weightings, see the documentation for Fuzzy Match Weightings.

Can be one of the following types:

The 'Constrained Wrap' property controls whether the extractor will match labels that wrap across multiple lines, but only when the wrap occurs at allowed positions (such as spaces or hyphens).

• Useful for documents where labels may be split across lines, but only in predictable ways.
• Helps match multi-line labels in tables or forms without introducing false positives from arbitrary line breaks.
• Configure minimum and maximum character or line counts to further restrict matching.

Example:
A table header like "Date of
Service" can be matched as "Date of Service" if constrained wrap is enabled.

See Constrained Wrap Options for more configuration details.

Can be one of the following types:

The 'Vertical Wrap' property allows the extractor to match multi-word labels that are stacked vertically, with each word on a separate line.

• Useful for forms or tables where labels are presented in a vertical arrangement.
• When enabled, the extractor can recognize labels such as "Purchase
  Order
  Number" as a single logical label.
• Additional options allow you to permit or restrict horizontal or vertical lines between text segments.

Example:
A label written as: > Purchase
> Order
> Number will be matched as "Purchase Order Number" if vertical wrap detection is enabled.

See Vertical Wrap Detection for more configuration options.