Parsing Complex Tables in Documents: Lettria’s Approach

Introduction

Extracting structured data from documents with complex layouts is a persistent challenge, especially when it comes to tables. These layouts often contain nested tables, merged cells, and irregular grids that disrupt standard parsing methods. This guide outlines Lettria’s approach to parsing complex tables using model-based predictions, improving accuracy and reliability in document intelligence workflows.

1. What Makes Tables with Complex Layouts Difficult to Parse?

Real-world documents often include tables that go far beyond simple rows and columns. Cells may span multiple rows or columns, tables may be nested within others, and visual alignment may not correspond to logical relationships. These elements introduce ambiguity that affects downstream data extraction, creating gaps or errors in interpretation. Without a precise understanding of the table’s structure, it’s difficult to automate or scale information retrieval.

2. Why Standard Parsing Tools Struggle with Complex Tables

Most traditional parsers rely on grid assumptions or fixed-layout heuristics. They segment based on visual lines or bounding boxes, which quickly break down when tables deviate from regular shapes. Merged or nested cells often lead to duplicate or fragmented outputs. These tools lack the flexibility to adapt to irregular real-world layouts, especially in documents like insurance claims, financial reports, or clinical studies.

3. How Lettria Parses Complex Tables

Lettria uses predictions from Microsoft’s TATR (Table Transformer) model to detect and reconstruct complex table structures. This model-based approach focuses on understanding relationships between individual cells, rather than applying rigid rules. It handles merged cells, nested layouts, and irregular grids by interpreting the table as a set of logically linked elements. The result is a more accurate and adaptable parsing process compared to conventional methods.

4. The Role of the Hierarchical Tree Structure

While Lettria uses a hierarchical tree to represent document layout, this structure does not directly influence table parsing. The tree organizes all document elements—titles, paragraphs, figures, and tables—into a logical flow. This helps with contextual positioning and navigation but keeps the table parsing process modular and independent. This separation allows Lettria to maintain precision in tables without disrupting the overall document structure.

5. Benefits for Document Workflows

• Higher accuracy in extracting data from visually complex tables
• Preserves logical structure, reducing the need for manual correction
• Modular architecture: table parsing and document structure are handled independently
• Domain adaptability: works on multilingual documents and domain-specific layouts (legal, insurance, healthcare)

These advantages streamline data workflows, enabling automation in environments where accuracy and compliance are critical.

6. Real-World Example: Parsing Complex Insurance Claims Tables

Insurance claims often include tables with policy details, payment schedules, and incident records. These tables feature irregular cell spans and sometimes embed nested structures. Lettria applies TATR to detect these configurations and extract coherent tabular data, which can then feed into claim management systems. This reduces manual review time and improves compliance tracking across large volumes of claims.

7. Use Cases

• Insurance: Parsing detailed claims and policy tables
• Life sciences: Extracting data from clinical trial reports
• Finance: Handling tables in annual reports and disclosures
• Healthcare: Structuring multi-level patient data across forms

Each use case benefits from precise extraction and improved downstream automation.

Conclusion & Call to Action

Parsing complex tables cannot be solved with simple rule-based systems. Lettria’s integration of TATR predictions enables high-fidelity extraction even in documents with irregular structures. By separating table parsing from global document layout, Lettria ensures modularity, scalability, and consistency.

If your workflows depend on accurate table extraction in complex documents, request a tailored demo to explore how Lettria can help.