PDF Maximum Number Finder

Two Python tools to find the largest numerical value in a PDF document.

Overview

This package provides two approaches:

1. Basic Version (pdf_max_finder.py) - Extracts literal numbers as they appear
2. NLP-Enhanced Version (pdf_max_finder_nlp.py) - Uses natural language processing to understand context and scale modifiers

Key Difference

The NLP version understands contextual clues like:

• "Revenue was $3.5 million" → interprets as $3,500,000 (not $3.5)
• "Population in thousands: 250" → interprets as 250,000
• "GDP (in billions): 21.5" → interprets as 21,500,000,000
• Scientific notation: "1.5e10" → interprets as 15,000,000,000
• Power notation: "10^120" → interprets as 10^120
• Abbreviations: "150K", "3.5M", "2.1B" → scaled appropriately

Requirements

• Python 3.6 or higher
• PyPDF2 library
• (Optional, for NLP-enhanced mode) spaCy + English model

Installation

pip install PyPDF2

Or using requirements.txt:

pip install -r requirements.txt

If you want the NLP-enhanced features (recommended), install spaCy and an English model:

python -m pip install -r requirements.txt
# then download the spaCy English model
python -m spacy download en_core_web_sm

Usage

Basic Version (Literal Numbers Only)

python pdf_max_finder.py '.\test_documents\FY25 Air Force Working Capital Fund.pdf'

This finds the largest number as literally written in the document.

NLP-Enhanced Version (Context-Aware)

python pdf_max_finder_nlp.py '.\test_documents\FY25 Air Force Working Capital Fund.pdf'

This understands context and applies scale multipliers to find the true largest value.

Examples

Given a document with:

• "Revenue: $3.5 million"
• "Population: 450,000"
• "Budget in billions: 2.1"

Basic version would report: 450,000 (largest literal number)

NLP version would report: 2,100,000,000 (2.1 billion after applying context)

Output Comparison

Basic Version Output

Reading PDF: document.pdf
Processing 10 pages...
Extracted 50000 characters of text
Found 234 numerical values

==================================================
LARGEST NUMBER FOUND: 450,000.00
==================================================

NLP-Enhanced Version Output

Reading PDF: document.pdf
Processing 10 pages...
Extracted 50000 characters of text
Analyzing numbers with NLP context...
Found 234 numerical values (including scaled values)

======================================================================
LARGEST NUMBER FOUND (NLP-Enhanced): 2,100,000,000.00
Context: 2.1 (scaled by billion: x1,000,000,000)
======================================================================

Top 10 Largest Numbers Found:
----------------------------------------------------------------------
 1.      2,100,000,000.00  |  2.1 (scaled by billion: x1,000,000,000)
 2.      3,500,000.00      |  3.5 (scaled by million: x1,000,000)
 3.        450,000.00      |  450,000
...
----------------------------------------------------------------------

💡 Note: This uses NLP to understand context like 'in millions'
   and applies appropriate scale multipliers to find the true
   largest value in the document.

Supported Scale Indicators

The NLP version recognizes these scale modifiers:

Full Words:

• trillion, billion, million, thousand, hundred
• Their plural forms (trillions, billions, etc.)

Abbreviations:

• K (thousand), M (million), B (billion), T (trillion)
• mn, mil, bn, tn

Metric Prefixes:

• kilo, mega, giga

Contextual Patterns:

• "3.5 million dollars"
• "in millions"
• "(millions)"
• "millions of dollars"
• "revenue of $3.5M"

Scientific/Power Notation:

• 1.5e10 (scientific notation)
• 10^120 (power notation)

How the NLP Version Works

1. Text Extraction: Extracts all text from the PDF
2. Context Analysis: For each number found, analyzes surrounding text (50 characters before and after)
3. Scale Detection: Looks for scale modifier keywords near the number
4. Multiplier Application: Applies the appropriate multiplier to get the actual value
5. Comparison: Compares all scaled values to find the maximum

Limitations

NLP Version

• Context window is limited to 50 characters around each number
• May misinterpret ambiguous cases
• Assumes standard English scale terminology
• Very large exponents (>1000) are skipped to prevent overflow

Both Versions

• Require text-based PDFs (not scanned images)
• Cannot process PDFs that require authentication
• Phone numbers and dates may be interpreted as numbers

Which Version Should I Use?

Use the Basic Version when:

• You want the largest number as literally written
• The document doesn't use scale modifiers
• You need simple, straightforward extraction

Use the NLP Version when:

• Documents express values "in millions/billions"
• You need to understand the true economic/scientific values
• Financial reports, research papers, or statistical documents
• Documents use abbreviations like "150K" or "3.5M"

Testing

# Basic version
python pdf_max_finder.py '.\test_documents\FY25 Air Force Working Capital Fund.pdf'

# NLP version  
python pdf_max_finder_nlp.py '.\test_documents\FY25 Air Force Working Capital Fund.pdf'

Error Handling

Both programs handle:

• Missing files
• Corrupted PDFs
• Documents with no text
• Documents with no numbers
• Invalid number formats

License

Open source - feel free to use and modify as needed.