Axiom vs Generic OCR

Most OCR tools were built to read documents.
Mathematics is not a document.

This page explains the fundamental technical difference between generic OCR systems and Axiom—and why that difference determines whether handwritten STEM notes become usable or unusable.

What generic OCR is designed for

The original purpose of OCR

Generic OCR systems (such as document scanners, camera OCR, and PDF converters) are optimized for:

• Paragraphs of prose
• Printed fonts
• Linear reading order
• Single-baseline text

They excel at:

• Books
• Contracts
• Articles
• Typed notes

In these contexts, OCR performs exactly as intended.

Why that breaks for math

Mathematics is not linear text

Handwritten mathematics relies on:

• Vertical relationships
• Nested hierarchies
• Alignment across multiple baselines
• Spatial grouping

Examples:

• Superscripts change meaning based on position
• Fractions require vertical parsing
• Matrices depend on row and column alignment
• Limits sit above and below operators

Generic OCR does not model these relationships.

It reads symbols in sequence, not in structure.

Typical failure modes of generic OCR

What actually goes wrong

When applied to handwritten math, generic OCR commonly produces:

• Flattened fractions rendered as inline text
• Superscripts misread as adjacent characters
• Matrices collapsed into unreadable sequences
• Loss of alignment in multi-line equations
• Dropped Greek symbols or operators
• Output that looks digital but cannot compile

The result often requires more manual correction than retyping the equation from scratch.

How Axiom approaches the same input

Structural parsing, not character scanning

Axiom does not treat math as text.

It analyzes handwritten pages as spatial systems, identifying:

• Baselines and vertical offsets
• Group boundaries such as fractions, radicals, and matrices
• Symbol roles based on position and context
• Relationships between symbols across space

Only after this structural understanding is established does Axiom generate output.

Where generic OCR still makes sense

This is not a blanket rejection

Generic OCR is appropriate for:

• Typed documents
• Prose-heavy notes
• Forms and contracts
• Simple handwritten text

Axiom is not designed to replace these tools.

It exists because STEM notation breaks them.

Where Axiom is the correct tool

Use Axiom when:

• You write equations by hand
• You work with matrices, integrals, or physics notation
• You need LaTeX or Markdown output
• You care about correctness, not screenshots
• You plan to reuse the content in formal academic or technical work

Axiom is optimized for STEM handwriting, not general text.

Overleaf is not an alternative

Axiom vs Overleaf

Overleaf is a LaTeX editor.
Axiom is a LaTeX generator.

Typing complex equations into Overleaf:

• Requires LaTeX syntax knowledge
• Is slow and error-prone
• Interrupts the natural flow of handwritten work

Axiom removes the manual coding step.

Write naturally. Convert once. Edit digitally.

Accuracy is not about “AI”

Why buzzwords don’t matter here

Accuracy in math recognition does not come from:

• Larger generic models
• Better cameras
• More post-processing

It comes from:

• Structural understanding
• Domain-specific training
• Respect for mathematical hierarchy

This is why generic OCR fails consistently—and why Axiom exists as a separate category.