Last updated:
Stylometry Lab
Compare two or more writing samples to estimate whether they share an author. The lab computes Burrows' Delta over the most-frequent words and projects the texts with PCA, so stylistically similar documents cluster together. The engine is real Python (NumPy) compiled to WebAssembly and runs entirely in your browser — your text is never uploaded.
How stylometry attribution works
Authors leave a fingerprint in how often they use ordinary function words — the, of, and, whilst, gonna. Burrows’ Delta turns each text into a vector of z-scored most-frequent-word frequencies and measures the distance between them; smaller distances suggest shared authorship. PCA projects those vectors into two dimensions so you can see clusters. It works best with several hundred words per sample and known comparison texts, and it is an investigative aid, not courtroom proof.
Frequently asked questions
What is Burrows' Delta?
A standard stylometry measure. Each text becomes a vector of z-scored frequencies of the most common words; the Delta distance is the mean absolute difference between two such vectors. Lower Delta means more similar style, which can indicate shared authorship.
How much text do I need?
More is better. A few hundred words per sample is a practical minimum; very short texts produce unstable results. Provide known reference samples plus the questioned text for the most useful comparison.
Is my text uploaded?
No. The analysis runs in your browser through Python (NumPy) compiled to WebAssembly. After the one-time engine download it works offline and your text never leaves your device.
Can this prove who wrote something?
No. Stylometry produces probabilistic leads, can be confounded by topic, translation, editing, or deliberate obfuscation, and should be corroborated. Treat it as one signal among many, not proof.