👁 Image
FieldCure.DocumentParsers 2.0.1

FieldCure.DocumentParsers

Lightweight document text extraction for .NET — DOCX, HWPX, XLSX, PPTX, HTML, and PDF. Structured Markdown output for LLM / RAG consumption. Pure managed, no native binaries.

Features

• DOCX — Headings, paragraphs, tables (nested), math → LaTeX, metadata → YAML, footnotes, endnotes, comments, headers/footers
• HWPX — Korean standard (KS X 6101 / OWPML). Headings, paragraphs, tables, math → LaTeX, metadata → YAML, footnotes, endnotes, memos, headers/footers
• XLSX — Sheets as markdown tables, metadata → YAML
• PPTX — Slide text, tables, speaker notes, metadata → YAML
• HTML — Readable content extraction via SmartReader → GitHub-flavored Markdown via ReverseMarkdown
• PDF — Page-by-page text extraction via PdfPig with ## Page {n} headers (auto-registered)
• Math equations — DOCX (m:oMath) and HWPX (hp:equation) converted to [math: LaTeX]
• Metadata — YAML front matter (title, author, created, modified, subject, keywords, description)
• Footnotes / Endnotes — [^N] / [^enN] inline references with definition sections
• Comments — Inline blockquote > **[Comment — author]:** format
• Factory pattern — DocumentParserFactory.GetParser(".docx") returns the right parser
• Zero platform dependency — Targets net8.0 and net10.0, no native binaries, no Windows-specific APIs
• Extensible — Implement IDocumentParser and call DocumentParserFactory.Register()

Install

dotnet add package FieldCure.DocumentParsers

Quick Start

using FieldCure.DocumentParsers;

// Auto-detect parser by extension — PDF is registered out of the box.
var parser = DocumentParserFactory.GetParser(".pdf");
if (parser is not null)
{
 var bytes = File.ReadAllBytes("report.pdf");
 var text = parser.ExtractText(bytes);
 Console.WriteLine(text);
}

// Check all supported extensions
foreach (var ext in DocumentParserFactory.SupportedExtensions)
 Console.WriteLine(ext); // .docx, .hwpx, .xlsx, .pptx, .html, .htm, .pdf

// Opt-out control for metadata, footnotes, etc.
var docxParser = new DocxParser();
var options = new ExtractionOptions
{
 IncludeMetadata = false,
 IncludeFootnotes = false
};
var text = docxParser.ExtractText(File.ReadAllBytes("report.docx"), options);

Output Format

Headings are prefixed with # markers. Tables are rendered as markdown. Documents with metadata include YAML front matter; footnotes/endnotes are rendered as reference-style links:

---
title: 2026 Business Plan
author: Alice
created: 2026-04-01
---

> **[Header]:** Company Confidential

# 2026 Business Plan

Please refer to the table below[^1] for details.

| Category | Q1 | Q2 |
| --- | --- | --- |
| Revenue | 100 | 150 |
| Cost | 80 | 90 |

> **[Footer]:** Page 1

## Footnotes

[^1]: Source: internal finance report.

Pipe characters inside cells are escaped as \| to preserve table structure. Use ExtractionOptions to selectively disable metadata, footnotes, comments, or headers/footers.

Supported Formats

Optional PDF Extensions

PDF text extraction is built in. Two sibling packages add extra PDF capabilities that require native binaries:

• Page image rendering — FieldCure.DocumentParsers.Imaging adds PdfImageRenderer : IMediaDocumentParser (PDFium via PDFtoImage).
• OCR fallback for scanned PDFs — FieldCure.DocumentParsers.Ocr adds OcrPdfParser + TesseractOcrEngine (English + Korean tessdata).

// Imaging (page → PNG)
using FieldCure.DocumentParsers.Imaging;
DocumentParserFactoryImagingExtensions.AddImagingSupport();

// OCR (scanned PDFs)
using FieldCure.DocumentParsers.Ocr;
using var ocr = DocumentParserFactoryOcrExtensions.AddOcrSupport();

Related Packages

• FieldCure.DocumentParsers.Imaging — PDF page image rendering (PDFium)
• FieldCure.DocumentParsers.Ocr — Tesseract OCR fallback for scanned PDFs
• FieldCure.AssistStudio.Core — AI provider client library that uses this package for document attachments

License

MIT — Copyright (c) 2026 FieldCure Co., Ltd.