15-docling

Docling Tutorials

This directory contains tutorials focused on Docling, IBM's open-source toolkit for parsing and converting documents from a wide range of formats.

What is Docling?

Docling is an IBM open-source library for parsing documents and exporting them to preferred formats. It supports:

• Input formats: PDF, DOCX, PPTX, XLSX, Images, HTML, AsciiDoc, Markdown
• Output formats: Markdown, JSON
• OCR support: Optical character recognition for scanned documents

Common use cases include processing medical records, banking documents, travel documents, and any scenario where unstructured data needs to be converted into a machine-readable structured format.

Prerequisites

Each tutorial in this directory includes its own setup and installation instructions. Please refer to the individual tutorial files for specific requirements.

Common requirements:

• Python 3.10 - 3.13
• IBM watsonx.ai account

Quick Start

1. Install dependencies (see Installation above)
2. Navigate to this directory:

   cd tutorials/15-docling

3. Open and run the tutorials

Tutorials

1. Convert Unstructured Data to Structured Data (unstructured-data-conversion.ipynb)

Use Docling with Python to convert unstructured data contained in a group of scanned files into a structured format.

• Author: Erika Russi
• Topics: Document parsing, unstructured-to-structured conversion, OCR, Docling export formats
• Prerequisites: RAG dependencies (requirements-rag.txt)
• Estimated time: 30–40 minutes

---

Related Docling Tutorials

The following Docling tutorials are located in the 01-rag-and-retrieval category:

Docling Granite Question Answering

Build a document-based question answering system using Docling and IBM Granite 3.1.

• Topics: Document parsing, Q&A systems, Docling + Ollama integration, large context windows
• Prerequisites: RAG dependencies + Ollama

DeepSeek RAG Reasoning with Docling

Advanced RAG with reasoning capabilities using DeepSeek and Docling on watsonx.

• Topics: Reasoning over documents, advanced retrieval, Docling integration, watsonx
• Prerequisites: RAG dependencies + Ollama

---

Key Concepts

Structured vs. Unstructured Data

• Structured data: Information organized into a fixed field within a record or file (SQL databases, JSON, XML, CSV, Excel spreadsheets). Ready for efficient data processing, analysis, and management.
• Unstructured data: Information that does not conform to a predefined data model or schema. Typically text-heavy (emails, social media posts, customer reviews) or non-text formats (audio recordings, video files, images). Makes up ~90% of enterprise information.

Why Convert Unstructured to Structured?

• Analysis automation: Run real-time queries and analytics
• Machine interpretability: Algorithms can process structured data directly
• Integration: Feed into databases, pipelines, and downstream AI systems

Docling Pipeline

1. Document Loading: Ingest files (PDF, DOCX, images, etc.)
2. Parsing: Extract text, tables, and layout information
3. OCR (optional): Recognize text in scanned images
4. Export: Output as Markdown or JSON for downstream use

Common Use Cases

• Healthcare: Scan and structure medical records and lab reports
• Finance: Process banking documents, invoices, and statements
• Legal: Extract structured data from contracts and filings
• Logistics: Parse shipping documents and travel records
• RAG pipelines: Pre-process documents before embedding and retrieval

Additional Resources

• Docling GitHub Repository
• IBM Think: Unstructured Data
• IBM Think: Structured vs. Unstructured Data
• IBM Think: OCR
• Main Repository README
• RAG and Retrieval Tutorials

Contributing

Found an issue or want to add a new Docling tutorial? Please open an issue or pull request in the GitHub repository.

License

See the LICENSE file in the repository root for license information.