TextLayout

Extract structured text from documents with columnar layouts.

Overview

TextLayout uses an XY-Cut algorithm to detect text blocks in documents that have been extracted from PDFs or other sources with preserved character spacing. It's particularly effective for:

• Invoices with multiple columns
• Forms with label:value pairs
• Documents with side-by-side content
• Any text where spatial positioning matters

Installation

pip install git+https://github.com/asynkron/Asynkron.TextLayout.git

Usage

Command Line

textlayout document.txt
textlayout document.txt 3  # min_gap=3 for tighter column detection
textlayout document.pdf 2  # requires pdftotext on PATH

Python API

from textlayout import extract

# Read your document
with open("invoice.txt") as f:
    text = f.read()

output = extract(text, min_gap=2)
print(output)

PDF via pdftotext

Requires Poppler's pdftotext available on your PATH.

Install options:

• macOS: brew install poppler
• Debian/Ubuntu: sudo apt-get install poppler-utils
• Python wrapper: pip install pdftotext (https://pypi.org/project/pdftotext/)

from textlayout import extract_pdf

output = extract_pdf("invoice.pdf", min_gap=2)
print(output)

How It Works

1. Text to Matrix: Converts text into a 2D character grid
2. Horizontal Split: Divides document into sections at blank lines
3. Vertical Split: Divides each section into columns at whitespace gaps
4. Normalization:
   • Joins label: with following value
   • Unwraps word-wrapped lines
   • Pulls up numbers after separators
5. Formatting:
   • Collapses multiple blank lines
   • Aligns key:value pairs in groups

Example

Input (raw PDF text with spacing):

Kund nr             Fakturanr          Fakturadatum
4601691270          005597910          2021-12-03

Output:

Kund nr     : 4601691270
Fakturanr   : 005597910
Fakturadatum: 2021-12-03

License

MIT