Image Text Finder
How to Detect Text Similarity and Plagiarism in Images

How to Detect Text Similarity and Plagiarism in Images

Detecting similar or plagiarized text in scanned images is important for academic integrity, content review, and legal protection. Aspose.OCR Image Text Finder for .NET enables batch detection of content similarity across archives or document sets.

Real-World Problem

Manual detection of duplicate or copied text in scanned images is inefficient and error-prone. Automated OCR-driven comparison allows scale and repeatability for academic, business, or legal needs.

Solution Overview

Extract text from each image and compare it against a corpus or target set for high similarity. Flag and report suspected matches for review or follow-up.

Prerequisites

  1. Visual Studio 2019 or later
  2. .NET 6.0 or later (or .NET Framework 4.6.2+)
  3. Aspose.OCR for .NET from NuGet
PM> Install-Package Aspose.OCR

Step-by-Step Implementation

Step 1: Prepare Your Image Sets 

string[] archiveFiles = Directory.GetFiles("./archive", "*.png");
string[] submissionFiles = Directory.GetFiles("./submissions", "*.png");

Step 2: Extract Text from Images 

RecognitionSettings settings = new RecognitionSettings();
settings.Language = Language.English;
AsposeOcr ocr = new AsposeOcr();
Dictionary<string, string> archiveTexts = new Dictionary<string, string>();
foreach (string file in archiveFiles)
{
    string text = ocr.Recognize(new OcrInput(InputType.SingleImage) { file }, settings)[0].RecognitionText;
    archiveTexts[file] = text;
}

Step 3: Compare for Similarity or Duplication

Use a simple text similarity function (e.g., Levenshtein distance, Jaccard index) or a .NET package for fuzzy matching:

foreach (string subFile in submissionFiles)
{
    string subText = ocr.Recognize(new OcrInput(InputType.SingleImage) { subFile }, settings)[0].RecognitionText;
    foreach (var kvp in archiveTexts)
    {
        double similarity = JaccardSimilarity(subText, kvp.Value); // custom function or library
        if (similarity > 0.8) // Tune threshold for your needs
        {
            File.AppendAllText("plagiarism_log.csv", $"{subFile},{kvp.Key},{similarity}\n");
        }
    }
}

// Example Jaccard similarity (token-based)
double JaccardSimilarity(string text1, string text2)
{
    var set1 = new HashSet<string>(text1.Split());
    var set2 = new HashSet<string>(text2.Split());
    int intersect = set1.Intersect(set2).Count();
    int union = set1.Union(set2).Count();
    return (double)intersect / union;
}

Step 4: Log and Review Results

  • Export suspected matches for human or academic/legal review

Step 5: Complete Example 

using Aspose.OCR;
using System;
using System.Collections.Generic;
using System.IO;

class Program
{
    static void Main(string[] args)
    {
        string[] archiveFiles = Directory.GetFiles("./archive", "*.png");
        string[] submissionFiles = Directory.GetFiles("./submissions", "*.png");
        RecognitionSettings settings = new RecognitionSettings();
        settings.Language = Language.English;
        AsposeOcr ocr = new AsposeOcr();
        Dictionary<string, string> archiveTexts = new Dictionary<string, string>();
        foreach (string file in archiveFiles)
            archiveTexts[file] = ocr.Recognize(new OcrInput(InputType.SingleImage) { file }, settings)[0].RecognitionText;
        foreach (string subFile in submissionFiles)
        {
            string subText = ocr.Recognize(new OcrInput(InputType.SingleImage) { subFile }, settings)[0].RecognitionText;
            foreach (var kvp in archiveTexts)
            {
                double sim = JaccardSimilarity(subText, kvp.Value);
                if (sim > 0.8)
                    File.AppendAllText("plagiarism_log.csv", $"{subFile},{kvp.Key},{sim}\n");
            }
        }
    }
    static double JaccardSimilarity(string text1, string text2)
    {
        var set1 = new HashSet<string>(text1.Split());
        var set2 = new HashSet<string>(text2.Split());
        int intersect = set1.Intersect(set2).Count();
        int union = set1.Union(set2).Count();
        return (double)intersect / union;
    }
}

Use Cases and Applications

Academic Integrity and Plagiarism Detection

Screen student submissions for copied content against archived sources.

Legal and Contract Review

Detect reuse or copying of contractual language in legal scanned docs.

Content Publishing and Media

Identify duplication or unauthorized reuse of text in creative industries.

Common Challenges and Solutions

Challenge 1: OCR Recognition Errors

Solution: Use high-quality scans and tune similarity thresholds.

Challenge 2: Large Archive Sets

Solution: Pre-index or batch process, parallelize if needed.

Challenge 3: Language or Formatting Variations

Solution: Normalize text (lowercase, remove stopwords), process per language set.

Performance Considerations

  • Text similarity calculations are compute-intensive for big sets—batch and schedule
  • Log all results for review and audit

Best Practices

  1. Validate flagged results with manual or committee review
  2. Tune similarity thresholds for accuracy vs. false positives
  3. Archive all logs for compliance and audit
  4. Use structured text normalization

Advanced Scenarios

Scenario 1: Visualize Similarity Results

Create charts or graphs from your CSV using Excel or BI tools.

Scenario 2: API Integration for Real-Time Submission Screening

Screen images upon upload and provide instant similarity feedback.

Conclusion

Aspose.OCR Image Text Finder for .NET empowers scalable, automated detection of similar or plagiarized content in images—essential for academia, legal, and publishing workflows.

See Aspose.OCR for .NET API Reference for more advanced comparison and search APIs.

 English