This document provides an overview of the three distinct content detection levels available in the Copyleaks API: Identical Matches, Minor Changes, and Paraphrased Content (Related Meaning). Understanding these detection levels helps you configure your content authenticity scans for optimal results.

📖 Introduction

Copyleaks offers a sophisticated, multi-layered approach to content matching that goes beyond simple word-for-word comparison. By leveraging different detection levels, the system can identify various forms of content reuse, from exact duplication to clever paraphrasing and even cross-language translation.

Each detection level serves a specific purpose and can be enabled or disabled according to your specific needs:

1. Identical Matches: Exact word-for-word matches
2. Minor Changes: Content with slight variations
3. Paraphrased Content: Rewritten text that conveys the same meaning

🔍 Detection Levels Explained

Identical Matches

Identical matches represent the most straightforward type of content matching, focusing on exact, word-for-word duplication.

• Description: Detects content that has been copied verbatim from another source without any alterations.
• Use Case: Ideal for finding direct plagiarism where content has been copied and pasted without modification.
• Configuration: Set properties.filters.identicalEnabled to true (enabled by default).
• Example:
  • Original: “The quick brown fox jumps over the lazy dog.”
  • Matched: “The quick brown fox jumps over the lazy dog.”

Tip

For cases where you only want to detect identical matches, you can disable the other detection levels. See our Identical Matches Detection guide for details.

Minor Changes

The minor changes detection level identifies content that has been slightly modified but remains fundamentally the same as the original source.

• Description: Detects content where small changes have been made, such as altering word forms, switching tenses, or making minimal substitutions.
• Use Case: Helpful for identifying cases where someone has made superficial changes to disguise copied content.
• Configuration: Set properties.filters.minorChangesEnabled to true (enabled by default).
• Examples:
  • Original: “The quick brown fox jumps over the lazy dog.”
  • Matched: “A quick brown fox jumps over the lazy dog.”

Paraphrased Content (Related Meaning)

The paraphrased content detection level identifies substantial rewrites that maintain the same core meaning as the original source.

• Description: Detects content that has been significantly rewritten while preserving the original meaning or ideas.
• Use Case: Essential for identifying sophisticated plagiarism where content has been carefully reworded to avoid detection.
• Configuration: Set properties.filters.relatedMeaningEnabled to true (enabled by default).
• Examples:
  • Original: “The quick brown fox jumps over the lazy dog.”
  • Matched: “That speedy brown fox just jumped right over a sleeping dog” (same meaning, different words)

Note

The paraphrased content detection uses advanced natural language processing to understand semantic meaning beyond simple word matching.

Cross-Language Detection

A particularly powerful feature of Copyleaks’ paraphrased content detection is its ability to identify content that has been translated from one language to another.

• Description: Detects content that has been translated from the original source into a different language.
• Use Case: Critical for organizations working in multilingual environments or checking content across language boundaries.
• Configuration: Configure through the properties.scanning.crossLanguages property.
• Example:
  • Original (English): “The quick brown fox jumps over the lazy dog.”
  • Matched (Spanish): “El rápido zorro marrón salta sobre el perro perezoso.”
  • Matched (French): “Le renard brun rapide saute par-dessus le chien paresseux.”

Caution

Cross-language detection requires additional processing and may consume additional credits. Only languages listed in the supported cross-languages documentation can be used.

🔧 Configuration

All three detection levels are enabled by default. You can customize which levels to include in your scan by modifying the properties.filters object in your API request:

Detection Levels Configuration

{
  "properties": {
    "filters": {
      "identicalEnabled": true,      // Enable/disable identical matches
      "minorChangesEnabled": true,   // Enable/disable minor changes detection
      "relatedMeaningEnabled": true  // Enable/disable paraphrased content detection
    }
  }
}

Recommended Configurations

Depending on your use case, you might want to adjust which detection levels are enabled:

High Precision (Fewer False Positives)

{
  "properties": {
    "filters": {
      "identicalEnabled": true,
      "minorChangesEnabled": true,
      "relatedMeaningEnabled": false
    }
  }
}

High Recall (Catch All Potential Matches)

{
  "properties": {
    "filters": {
      "identicalEnabled": true,
      "minorChangesEnabled": true,
      "relatedMeaningEnabled": true
    }
  }
}

Identical Only (Exact Matches Only)

{
  "properties": {
    "filters": {
      "identicalEnabled": true,
      "minorChangesEnabled": false,
      "relatedMeaningEnabled": false
    }
  }
}

🚀 Next Steps

After understanding the different detection levels:

Configure Identical Matches

Detect Plagiarism in Text

Detect AI-Generated Content

Cross-Language Detection

💬 Support

If you need assistance with configuring detection levels or have questions about which settings are best for your use case, please contact our support team.