AI Column Extractor

Extracts structured content from documents with two-column layouts.

Remarks

Overview

The AI Column Extractor is designed to extract structured data from documents that use a two-column layout. It leverages AI-driven layout analysis and OCR results to identify and extract content from key regions of each page, including the header, left column, right column, and footer.

Key Features

• Automatically detects and separates content into header, left column, right column, and footer regions.
• Uses AI to determine layout boundaries based on visual structure and spatial patterns.
• Integrates OCR results to extract text from each region.
• Provides diagnostic tools to visually confirm extracted regions and troubleshoot extraction issues.

When to Use

• Ideal for documents with a consistent two-column format, such as academic transcripts, reports, or structured forms.
• Not suitable for documents without a clear two-column layout, as the extractor may return no results.

Configuration Guidance

• Ensure documents are scanned with good quality and follow a clear two-column structure.
• Use the CustomInstructions property to provide document-type-specific extraction goals, such as identifying unique header, footer, or column content.

Diagnostics

• Diagnostic logging is supported to help users understand and troubleshoot the extraction process.
• When enabled, the extractor can generate visual annotations of bounding rectangles, a chat log of the AI interaction, and a schema file describing the expected layout.

Tips for Best Results

• Preprocess documents for clear formatting and alignment.
• Use diagnostic tools to verify the accuracy of extracted regions and data.

Properties

Name	Type		Description
Generator	Data Generator	►	Defines the generative AI settings used for LLM-powered extraction. EditorObject Properties Editor ConverterExpandable Converter The 'Generator' property configures how Grooper uses large language models (LLMs) to extract structured data from unstructured document content. Select a Data Generator to control the model, output format, and chat parameters for AI-driven extraction. The generator ensures schema-driven, reliable extraction and provides diagnostic logging for troubleshooting and auditing. Adjust generator settings to optimize accuracy, consistency, and performance for your specific use case.
Custom Instructions	String	►	Optional instructions for tailoring layout extraction to specific document types. EditorText Property Editor By default, the system uses built-in guidance to detect header, footer, and column regions. These general instructions are designed to work across a wide range of document formats without requiring customization. Use the CustomInstructions property if you need to provide additional, document-specific guidance to the AI. For example, you might use CustomInstructions to describe the kinds of content that typically appear in the header, footer, or column areas — such as expected labels, terminology, or patterns — to help the AI better distinguish between regions. These instructions are added to the prompt sent to the AI model, allowing it to more accurately interpret and adapt to your specific document type.

See Also

Data Generator

Used By

Document TypeExtract FromData ColumnData FieldLexicalRules-BasedSpell CorrectorAuto Complete SettingsParagraph MarkerMetadata OptionsOCR LayerLine Periodicity DetectorFixed WidthLabeled ValueSelect PageData TypeOCR ReaderDividerAnchorSimple