Document Tagging

​

Overview

​

Tag Structure

• Name: The identifier for the tag (e.g., ‘name’, ‘address’, ‘phone’)
• Value: The actual content being tagged
• Start/End Positions: Optional positions within the node’s content (if tagging specific text)
• UUID: Unique identifier for the tag instance
• Confidence: A score between 0 and 1 indicating tagging certainty
• Group UUID: Links related tags together
• Data: Additional JSON-serializable metadata
• Owner URI: Identifies the source that created the tag (e.g., a model reference)

​

Tagging Methods

​

1. Basic Node Tagging

# Basic node tagging
node.tag('category')

// Basic node tagging
await node.tag('category');

​

2. Fixed Position Tagging

# Tag text from position 6 to 12
node.tag('name', fixed_position=[6, 12])

​

3. Regular Expression Tagging

# Tag all email addresses in the content
node.tag('email', content_re=r'[\w\.-]+@[\w\.-]+')

​

Advanced Tagging Features

​

Tag Groups

import uuid

group_id = str(uuid.uuid4())
# Tag multiple related elements with the same group UUID
node.tag('person_name', fixed_position=[0, 10], group_uuid=group_id)
node.tag('person_age', fixed_position=[15, 17], group_uuid=group_id)

// Tag with group ID
await node.tagWithOptions('person_name', {
    start: 0,
    end: 10,
    groupId: 1
});
await node.tagWithOptions('person_age', {
    start: 15,
    end: 17,
    groupId: 1
});

​

Tag Metadata

node.tag('address', data={
    'type': 'residential',
    'verified': True
})

​

Tag Confidence

node.tag('product_code', confidence=0.95)

await node.tagWithOptions('product_code', {
    confidence: 0.95
});

​

Tag Owner URI

node.tag('invoice_number', owner_uri='model://kodexa/invoice-extractor')

​

Working with Tagged Content

​

Retrieving Tags

# Get all tags on a node
tags = node.get_tags()

# Iterate over tags
for tag in tags:
    print(f"Name: {tag.uuid}, Value: {tag.value}, Confidence: {tag.confidence}")

# Check if a node has a specific tag
if node.has_tag('address'):
    print("Node has address tag")

# Get tag names
tag_names = node.get_tag_names()

// Get all tag names on a node
const tags = await node.getTags();

// Check if a node has a specific tag
if (await node.hasTag('address')) {
    console.log("Node has address tag");
}

​

Removing Tags

# Remove a specific tag by name
node.remove_tag('category')

// Remove a specific tag by name
await node.removeTag('category');

​

Tag Instances

# Create a tag instance spanning multiple nodes
nodes = document.select('//line')
document.add_tag_instance('address_block', nodes)

​

Finding Tagged Nodes

# Find all nodes with a specific tag
tagged_nodes = document.select("//*[hasTag('company_name')]")

# Get all nodes with a specific tag
all_tagged = document.select("//*[hasTag()]")

// Find all nodes with a specific tag
const taggedNodes = await document.select("//*[hasTag('company_name')]");

// Get all tagged nodes
const allTagged = await document.getTaggedNodes();

​

Diagrams

​

Basic Tag Structure

​

Tag Relationships

​

Best Practices

1. Use Meaningful Tag Names: Choose descriptive names that reflect the content being tagged.
2. Group Related Tags: Use group_uuid (Python) or groupId (TypeScript) to group related pieces of information.
3. Include Confidence: When using automated tagging, include confidence scores.
4. Add Metadata: Use the data parameter to store additional context about the tag.
5. Set Owner URI: When tagging from models or automated processes, set the owner_uri to track the tag source.

​

Common Patterns

​

Document Classification

# Tag document type based on content
node.tag('document_type', value='invoice', data={
    'confidence': 0.98,
    'classifier': 'invoice_classifier_v1'
})

​

Entity Extraction

# Tag named entities using regex
node.tag('organization', content_re=r'Microsoft|Google|Apple')

​

Form Field Extraction

# Tag form fields with metadata
node.tag('field', fixed_position=[100, 150], data={
    'field_name': 'total_amount',
    'field_type': 'currency',
    'required': True
})

​

Tag Options Reference

​

Error Handling

1. Position Errors: Ensure fixed positions are within content bounds
2. Regular Expression Matching: Test patterns thoroughly
3. Node Selection: Verify node existence before tagging
4. Content Accessibility: Check content availability before tagging

# Example of safe tagging with error handling
try:
    if node.content:  # Check if content exists
        if len(node.content) >= end_position:  # Verify position
            node.tag('field', fixed_position=[start_position, end_position])
except Exception as e:
    print(f"Tagging error: {str(e)}")

​

Performance Considerations

1. Batch related tags together using group_uuid
2. Use specific selectors to limit the scope of tagging operations
3. Consider using tag instances for large groups of related nodes
4. Use transactions when performing many tag operations together