Transaction Extractor

Can be one of the following types:

Overview

Controls how the content of each transaction is formatted and presented to the AI model during extraction. The quoting method determines the structure, annotation, and delimiters used to provide transaction context, which can significantly impact extraction accuracy—especially for documents with complex or ambiguous layouts.

Configuration and Usage

• Select a quoting method that best preserves the structure and meaning of your transaction content.
• Quoting methods may add line numbers, delimiters, or special formatting to help the AI model understand the layout.
• Custom quoting methods can be used for specialized document types or extraction scenarios.

Impact

• The quoting method directly affects how the AI model interprets transaction boundaries and content.
• Consistent quoting improves the model’s ability to align extracted data with the original document.
• For multi-line or irregular transactions, use a quoting method that adds clear structure or annotations.

Example

For a table of invoice line items, use a quoting method that includes line numbers and column delimiters:

1 | Widget A | 10 | $5.00
2 | Widget B | 7  | $3.50

Can be one of the following types:

Overview

Specifies optional document-level content to be included in the extraction context for each transaction. This is useful for providing the AI model with information that appears outside of individual transactions, such as table headers, footers, or summary lines that are not repeated within each transaction.

Configuration and Usage

• Use this property to supply document content that should be considered when extracting data from transactions.
• Common examples include table headers, page-level notes, or section titles that provide context for all transactions.
• The selected quoting method determines how this content is formatted and annotated in the prompt.

Impact

• Including document-level context can improve extraction accuracy, especially when transactions rely on shared headers or labels.
• If left blank, only transaction content is provided to the AI model.

Example

To include a table header that appears once per page:

| Item | Quantity | Price |

The header will be quoted and presented before the list of transactions in the extraction prompt.

Overview

Allows you to provide custom instructions to the AI model, guiding how data should be extracted from each transaction. Use this property to clarify extraction requirements, handle edge cases, or provide document-specific rules that are not covered by the default schema.

Configuration and Usage

• Enter clear, concise instructions to improve extraction accuracy, especially for complex or non-standard documents.
• Instructions are appended to the prompt for each extraction operation.
• If left blank, the AI model will rely on the default schema and context.

Impact

• Well-crafted instructions can resolve ambiguities, enforce business rules, or direct the model to ignore irrelevant content.
• Overly broad or conflicting instructions may reduce extraction accuracy.

Example

Extract only the "Invoice Date" and "Total Amount" fields. Ignore any summary or footer lines.
Use the date format YYYY-MM-DD.

Best Practices

• Use this property to address ambiguous layouts, special formatting, or business rules not covered by the schema.
• Review extraction diagnostics to refine instructions as needed.

Use this property to limit extraction to a subset of fields, tables, or sections. This reduces prompt complexity and focuses the LLM on only the data you need.

• If left empty, all visible, non-computed descendant Data Elements are included.
• Add specific Data Fields, Data Tables, or Data Sections to target extraction and improve performance.
• Useful for incremental extraction or when only certain data is needed from a document.

Example:
To extract only the "Invoice Date" and "Total Amount" fields, add those fields to this list.

Best Practices

• Limit included elements to those required for your use case to improve extraction speed and accuracy.
• Use hierarchical selection to target deeply nested fields or tables as needed.

The 'Max Degree of Parallelism' property controls how many extraction operations (transactions) are processed simultaneously. Increasing this value can speed up extraction for large documents, but may increase resource usage.

Usage and Configuration

• Typical values range from 1 (sequential) to 32 (high parallelism).
• Adjust based on available system resources and desired performance.

Impact

• Higher values increase throughput but may consume more CPU and memory.
• Lower values reduce resource usage but may slow down processing.

Best Practices

• Start with the default value and adjust incrementally.
• Monitor system performance and adjust as needed to avoid overloading the environment.

The 'Transactions Per Operation' property determines how many transactions are grouped and processed together in a single extraction operation. Batching transactions can improve efficiency and reduce the number of LLM calls.

Usage and Configuration

• Set to a positive integer to control batch size.
• Typical values range from 1 (one transaction per operation) to higher values for bulk processing.

Impact

• Larger batch sizes reduce the number of LLM calls but may increase prompt complexity and risk exceeding context limits.
• Smaller batch sizes provide more granular control and may improve extraction accuracy for complex documents.

Best Practices

• Use smaller batches for documents with highly variable or complex transactions.
• Use larger batches for uniform, simple transactions to maximize throughput.