TXTextControl.Document.Classification

A .NET 10 console application that classifies .docx documents into predefined categories using weighted keyword profiles.

What this project does

The classifier:

• Loads classification profiles from classification-profiles.json.
• Extracts text from:
  • body paragraphs,
  • each section's headers and footers.
• Detects structure-aware regions (Title, Heading, Body, Header, Footer).
• Scores category rules using:
  • keyword match mode (WholeWord, Phrase, Exact),
  • keyword strength (Weak, Normal, Strong),
  • region weight,
  • position-aware weighting,
  • diminishing returns for repeated matches.
• Returns:
  • predicted category,
  • confidence,
  • per-category score details and keyword hits.

Tech stack

• Target framework: .NET 10 (net10.0)
• Main package: TXTextControl.TextControl.Core.SDK (34.0.3)

Project structure

• Program.cs - CLI entry point.
• Classification/DocxKeywordClassifier.cs - extraction + scoring pipeline.
• Classification/DocumentClassificationProfileLoader.cs - profile load/validation.
• Classification/Models/* - core model types.
• classification-profiles.json - category and rule configuration.
• Documents/ - sample .docx files.

Run the classifier

Pass the document path as the first argument.

cd TXTextControl.Document.Classification
dotnet run --project TXTextControl.Document.Classification.csproj -- "Documents/Sample Resume.docx"

Or from repo root:

dotnet run --project "TXTextControl.Document.Classification/TXTextControl.Document.Classification.csproj" -- "TXTextControl.Document.Classification/Documents/Sample Resume.docx"

Example output

Document: TXTextControl.Document.Classification/Documents/Sample Resume.docx
Classification: Resume
Confidence: 68.26%
Scores:
- Resume: 28.33
- Report: 8.74
- Financial: 7.11
...

Profile format (classification-profiles.json)

Each profile has a name and rules collection. Each rule supports:

• term (string)
• weight (number > 0)
• matchMode (WholeWord, Phrase, Exact)
• strength (Weak, Normal, Strong)
• optional: caseSensitive (bool)
• optional: targetRegion (Title, Heading, Body, Header, Footer)

Minimal example:

[
  {
    "name": "Resume",
    "rules": [
      {
        "term": "work experience",
        "weight": 3.0,
        "matchMode": "Phrase",
        "strength": "Strong"
      }
    ]
  }
]

Notes

• Only .docx files are supported.
• Profiles are validated at startup (duplicate categories/terms, empty terms, non-positive weights).
• If no profile scores above zero, result category is Unknown.