TorchV Unstructured

English | 中文

A powerful and developer-friendly document parsing library optimized for RAG (Retrieval Augmented Generation) applications. Built on top of industry-standard Java libraries like Apache Tika, Apache POI, and PDFBox, TorchV Unstructured provides enhanced parsing capabilities with intelligent table structure recognition and content extraction.

🚀 Key Features

• Intelligent Table Parsing: Advanced table structure analysis with proper cell merging detection
• Multi-format Support: Seamless handling of DOC, DOCX, PDF, and other document formats
• RAG-Optimized Output: Structured content extraction designed for AI/ML pipelines
• Markdown & HTML Export: Flexible output formats with preserved table structures
• Image Extraction: Automatic extraction and handling of embedded images
• Memory Efficient: Optimized for processing large documents with minimal memory footprint

📦 Installation

Maven

<dependency>
    <groupId>com.torchv.infra</groupId>
    <artifactId>torchv-unstructured</artifactId>
    <version>1.0.0</version>
</dependency>

Gradle

implementation 'com.torchv.infra:torchv-unstructured:1.0.0'

🔧 Quick Start

Basic Document Parsing

import com.torchv.infra.unstructured.UnstructuredParser;

// Parse document to markdown (recommended for RAG)
String content = UnstructuredParser.toMarkdown("document.docx");
System.out.println(content);

// Parse document to markdown with HTML tables (preserving table structure)
String contentWithTables = UnstructuredParser.toMarkdownWithHtmlTables("document.docx");
System.out.println(contentWithTables);

Advanced Table Extraction

import com.torchv.infra.unstructured.UnstructuredParser;

import java.io.File;
import java.util.List;

// Extract only tables from Word documents
List<String> tables = UnstructuredParser.extractTables("document.docx");
for (int i = 0; i < tables.size(); i++) {
    System.out.println("Table " + (i + 1) + ":");
    System.out.println(tables.get(i));
}

// Get structured result with more control
DocumentResult result = UnstructuredParser.toStructuredResult("document.docx");
if (result.isSuccess()) {
    System.out.println("Content: " + result.getContent());
    System.out.println("Tables: " + result.getTables());
}

File Format Support

import com.torchv.infra.unstructured.UnstructuredParser;
import com.torchv.infra.unstructured.util.UnstructuredUtils;

// Check if file format is supported
if (UnstructuredUtils.isSupportedFormat("document.docx")) {
    String content = UnstructuredParser.toMarkdownWithHtmlTables("document.docx");
    System.out.println("Parsing successful!");
} else {
    System.out.println("Unsupported file format");
}

// Get all supported formats
List<String> supportedFormats = UnstructuredUtils.getSupportedFormats();
System.out.println("Supported formats: " + String.join(", ", supportedFormats));

🎯 Core Components

Unified Entry Point

• UnstructuredParser: Main entry class providing simple and unified API for all document parsing operations

Document Parsers

• UnstructuredWord: Universal Word document parsing with auto-detection
• TikaAutoUtils: Generic document parsing with auto-detection (underlying implementation)
• WordTableParser: Specialized Word document table parser
• DocxTableParser: Advanced DOCX table structure analyzer

Content Handlers

• ToMarkdownWithHtmlTableContentHandler: Converts documents to Markdown with HTML tables
• DocMarkdownWithHtmlTableContentHandler: Specialized DOC format handler
• DocXMarkdownWithHtmlTableContentHandler: Specialized DOCX format handler

Table Analysis

• TableStructureAnalyzer: Intelligent table structure recognition
• CellMergeAnalyzer: Advanced cell merging detection
• HtmlTableBuilder: Clean HTML table generation

Utilities

• FileMagicUtils: File type detection and validation
• ImageExtractParse: Embedded image extraction

🔍 Advanced Usage

RAG Application Integration

import com.torchv.infra.unstructured.UnstructuredParser;
import com.torchv.infra.unstructured.core.DocumentResult;

// Optimized for RAG applications
public class RAGDocumentProcessor {

    public DocumentChunk processDocument(String filePath) {
        // Parse with table structure preservation for better context
        String content = UnstructuredParser.toMarkdownWithHtmlTables(filePath);

        // Extract tables separately for structured data processing
        List<String> tables = UnstructuredParser.extractTables(filePath);

        return new DocumentChunk(content, tables);
    }
}

Batch Processing

import com.torchv.infra.unstructured.UnstructuredParser;
import com.torchv.infra.unstructured.util.UnstructuredUtils;

public class BatchProcessor {

    public void processBatch(List<String> filePaths) {
        filePaths.parallelStream()
                .filter(UnstructuredUtils::isSupportedFormat)
                .forEach(this::processFile);
    }

    private void processFile(String filePath) {
        try {
            String content = UnstructuredParser.toMarkdownWithHtmlTables(filePath);
            // Save or further process the content
            saveProcessedContent(filePath, content);
        } catch (Exception e) {
            log.error("Failed to process file: {}", filePath, e);
        }
    }
}

Error Handling and Validation

import com.torchv.infra.unstructured.UnstructuredParser;
import com.torchv.infra.unstructured.util.UnstructuredUtils;

public class DocumentValidator {
    
    public ProcessingResult validateAndProcess(String filePath) {
        // Check file format
        if (!UnstructuredUtils.isSupportedFormat(filePath)) {
            return ProcessingResult.unsupportedFormat();
        }
        
        try {
            String content = UnstructuredParser.toMarkdownWithHtmlTables(filePath);
            List<String> tables = UnstructuredParser.extractTables(filePath);
            
            return ProcessingResult.success(content, tables);
        } catch (RuntimeException e) {
            return ProcessingResult.error(e.getMessage());
        }
    }
}

🌟 Why Choose TorchV Unstructured?

For RAG Applications

• Structured Output: Clean, structured content extraction perfect for embedding generation
• Table Preservation: Maintains table relationships crucial for document understanding
• Metadata Rich: Extracts comprehensive document metadata for enhanced retrieval

For Developers

• Simple API: Intuitive interfaces with sensible defaults
• Extensible: Plugin-based architecture for custom content handlers
• Production Ready: Battle-tested with comprehensive error handling

Performance Optimized

• Memory Efficient: Streaming processing for large documents
• Fast Processing: Optimized algorithms for quick parsing
• Scalable: Designed for high-throughput document processing

📚 Documentation

• API Documentation
• Examples Repository
• Migration Guide

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

🙏 Acknowledgments

• Apache Tika - Content analysis toolkit
• Apache POI - Java API for Microsoft Documents
• PDFBox - PDF document manipulation

📞 Support

• 📧 Email: xiaoymin@foxmail.com
• 🐛 Issues: GitHub Issues
• 💬 Discussions: GitHub Discussions

---

Made with ❤️ by TorchV Team