Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

Table extraction

Table extraction is the task of finding tables in documents and rebuilding their rows, columns, headers and cells as structured data, such as HTML, Markdown or CSV, instead of loose lines of text.

Also called: table structure recognition, table detection

By Divya Sharma · Checked Nov 15, 2026 · How we review

Prices, specifications, financial results and test data usually live in tables. A naive text extractor flattens a table into a run of numbers with no link to their headers, so a model asked "what was Q3 revenue?" cannot tell which figure belongs to which column. Table extraction keeps that structure, which makes it one of the harder and more valuable parts of document parsing.

How it works

The task is usually split in three. Table detection finds where tables are on the page, often as part of layout analysis. Structure recognition finds the rows, columns and spanning cells inside each table. Functional analysis identifies which cells are headers. Training data matters: PubTables-1M, a dataset of nearly one million tables from scientific articles, showed that transformer-based object detection models trained on it handle all three tasks well. The cell text comes from the PDF text layer or from OCR.

Example

An annual report has a table whose header spans two rows and whose first column groups line items by region. Good extraction outputs an HTML table with the merged header cells and region groups intact, so each number can be traced to its row and column. That output can then be stored whole, so chunking does not split a table across chunks.

Limits

Tables without ruling lines, tables that continue across pages, nested tables and charts that look like tables remain common failure cases. Check extracted numbers against the source before using them in reports.

Related terms

Document parsing
Document parsing converts files such as PDFs, scans, slides and spreadsheets into structured text that software can use, keeping headings, reading order, tables and images instead of a flat stream of characters.
Layout analysis
Layout analysis is the part of document parsing that finds the regions on a page, such as titles, paragraphs, tables, figures, captions and footers, labels each one, and works out the order a person would read them in.
Optical character recognition
Optical character recognition (OCR) is software that reads text from images, such as scanned pages, photos or screenshots, and turns it into machine-readable characters that can be searched, copied or fed to a model.
Chunking
Chunking is the step that splits long documents into smaller passages before they are embedded and indexed, so a retrieval system can find and return only the parts of a document that answer a query.
Structured output
Structured output is an API feature that makes a language model's reply follow a format you define, usually a JSON schema, so that code can parse it reliably instead of scraping free text.

Sources