Table Row Detector

Detects table rows in tabular data using value extractors and layout analysis.

Remarks

The Table Row Detector is a core component of the Tabular Layout extraction method, responsible for identifying and segmenting table rows within a document. It works in conjunction with the Table Header Detector and column configuration to accurately map document content into structured rows and columns.

How Row Detection Works

Row detection operates by analyzing the document's text lines and applying value extractors to locate column values. For a line to be considered a valid table row:

• It must contain values for all columns marked as required in Tabular Layout Options.
• It must meet or exceed the 'Minimum Cell Count' property, ensuring a minimum number of columns are present.
• The horizontal spacing between detected values must not exceed the 'Maximum Gap' property.
• The first row must appear within the distance specified by 'Maximum Header Distance' from the header.

The detector supports a wide range of table layouts, including:

• Tables with variable or missing columns.
• Multi-line rows (when used with Multiline Row Settings).
• Tables split across multiple regions or pages.

Configuration Guidance

• Minimum Cell Count:
  Increase this value to require more columns for a row to be detected, reducing false positives in noisy documents. Decrease for sparse or irregular tables.

• Maximum Gap:
  Adjust to control how far apart column values can be and still be grouped as a single row. Use higher values for wide tables or those with inconsistent spacing.

• Maximum Header Distance:
  Set to allow for extra space or blank lines between the header and the first row. Increase for documents with decorative or whitespace lines after the header.

• Find Column Positions:
  Enable to dynamically adjust header cell boundaries based on detected values, improving alignment in documents with shifting or misaligned columns.

• Merge Multiple Instances:
  Use to combine table regions that are separated by non-row content or page breaks into a single logical table.

Example Scenarios

• Invoice Line Items:
  Detects each line item as a row, even if some columns are occasionally missing (if not required).

• Multi-Page Tables:
  With 'Merge Multiple Instances' enabled, a table that continues across pages will be treated as a single table.

• Tables with Blank Lines:
  Increase 'Maximum Header Distance' to allow for blank or decorative lines between the header and first row.

Troubleshooting & Best Practices

• If extra or partial lines are detected as rows, increase 'Minimum Cell Count' or mark more columns as required.
• If rows are missed due to wide spacing, increase 'Maximum Gap'.
• Use diagnostic output to review which lines are detected as rows and adjust settings accordingly.
• Test with a variety of sample documents to ensure robust extraction across different layouts.

Related Concepts

• Tabular Layout: The extraction method that uses this detector.
• Table Header Detector: Configures header detection for the table.
• Tabular Layout Options: Per-column settings for required columns and detection rules.
• Data Table: The parent object representing the table structure.

For more information, see the documentation for Tabular Layout, Data Table, Table Header Detector, and Tabular Layout Options.

Properties

Name	Type		Description
Minimum Cell Count	Int32	►	The minimum number of column values required for a table row to be detected. Requiredtrue Default3 Min1 Controls how many columns must have detected values on a single line for it to qualify as a table row. This helps filter out lines that do not contain enough data to be considered valid rows. The line must also contain values for all columns marked as required in Tabular Layout Options. Increase this value to reduce false positives in documents with extra or partial lines. Example: If set to 3, a line must have values for at least three columns to be detected as a row.
Maximum Gap	Double	►	The maximum allowed gap between detected values, as a multiple of line height. ConverterPercent Converter Default0% Min0% Max800% Specifies how much horizontal space is permitted between detected column values within a row. This helps control whether widely separated values are grouped into the same row or treated as separate. A value of 1.0 allows a gap up to the average line height. Increase this value for tables with widely spaced columns; decrease for tightly packed tables. Example: For tables with double-spaced columns, set this value to 2.0.
Maximum Header Distance	Double	►	The maximum allowed distance from the header to the first detected row, as a multiple of line height. ConverterPercent Converter Default0% Min0% Max800% Determines how far below the header the first row can appear and still be considered part of the table. This is useful for handling tables with extra spacing or intervening content between the header and first row. A value of 1.0 allows the first row to be up to one line height below the header. Increase this value for tables with extra space after the header. Example: For tables with a blank line between the header and first row, set this value to 2.0.
Find Column Positions	Boolean	►	When true, adjusts header cell positions based on detected column values. DefaultFalse Enable this property to refine the positions of header cells using the actual detected values in the table rows. This can improve alignment between headers and values, especially in documents with inconsistent or misaligned layouts. Use when column values do not line up perfectly with header cells. May improve extraction accuracy for tables with variable formatting.
Merge Multiple Instances	Boolean	►	When true, merges multiple detected row spans into a single table instance. DefaultFalse Enable this property to combine multiple detected row regions into a single table, rather than treating them as separate tables. This is useful for documents where a table may be split across multiple regions or pages, but should be treated as a single logical table. Use for multi-page tables or tables interrupted by non-row content. Disable to keep separate table regions distinct.

Used By

Tabular Layout