PoetryScribe
Where meter meets machine
Windows
macOS
Linux
iOS
Android
Automated poetry scansion using Optimality Theory constraint evaluation—84% accurate at marking stressed and unstressed syllables. Works completely offline with a 134k-word pronunciation dictionary. No internet required, no data sent anywhere.
Why PoetryScribe is Different
High-Accuracy Scansion
PoetryScribe uses Optimality Theory multi-parse evaluation, generating ~10 candidate scansions and selecting the one with the fewest weighted constraint violations. Benchmarked at 84% syllable-level accuracy against expert annotations. You can refine any result manually.
100% Offline
PoetryScribe never sends your poetry anywhere. All analysis happens locally on your device. No cloud services, no API calls, no data collection. Works without internet after installation.
No Service Dependencies
Other tools rely on external services that can go offline, change pricing, or shut down. PoetryScribe will work forever - the software you download today will work exactly the same in 10 years.
Features
Automated Scansion
Mark stressed (/) and unstressed (˘) syllables using Optimality Theory constraint evaluation. Click any mark to toggle manually.
Rhyme Detection
Perfect rhymes, slant/near rhymes, assonance, and consonance via CMU dictionary phoneme matching.
Meter Identification
Detect iambic, trochaic, anapestic, dactylic patterns with confidence scoring and foot identification.
Synonym & Antonym Lookup
WordNet 3.0 + ConceptNet 5.7 semantic relations—explore word relationships while you write.
Sentiment Analysis
AFINN-165 valence scores + NRC emotion categories to analyze the emotional tone of your verse.
Fully Offline
All linguistic data ships with the app. No network required, no cloud dependencies, no service fees.
How It Works
Paste or type your poem into PoetryScribe. The app looks up each word in the CMU Pronouncing Dictionary, generates ~10 candidate scansions, and scores each against 9 weighted Optimality Theory constraints. The candidate with the fewest violations wins. Stressed syllables are marked with /, unstressed with ˘.
The algorithm achieves 84% accuracy against expert annotations from the University of Virginia's "For Better For Verse" corpus. You can click any stress mark to toggle it manually—your interpretation is final.
System Requirements
Windows
- Windows 10 or later
- 4 GB RAM minimum
- 100 MB disk space
macOS
- macOS 10.14 or later
- 4 GB RAM minimum
- 100 MB disk space
Linux
- Ubuntu 20.04+ or equivalent
- 4 GB RAM minimum
- 100 MB disk space