VoiceLab
A local writing-analysis prototype that keeps text on the device.
In 30 seconds
- Problem
- Can writing signals be explained, and a text revised, without losing what it means?
- What it is
- A local prototype for explainable writing-signal analysis and meaning-preserving revision. It explores patterns in language while keeping text on the local device.
- Stage
- Local prototype
Focus
Explainable analysis and thoughtful revision
Try the signals
This page shows the idea with sample text; the prototype itself isn’t shown here.
Every signal below is computed by this page, in your browser, as you type. The text is never sent anywhere.
It stays in this tab: nothing is sent or stored.
Words per sentence
8 · 19 · 11 · 6 · 26
- Average sentence
- 14.0 words
- Words per sentence. It shows rhythm, not quality.
- Word variety
- 0.76 53 of 70 words are distinct
- Distinct words divided by all words, the type/token ratio. It falls as a text gets longer, so compare texts of similar length.
- Long words
- 11% seven letters or more
- A hint of density. It says nothing about whether a word is the right one.
- Reading ease
- 75 fairly easy
- Flesch reading ease: 206.835 − 1.015 × (words ÷ sentences) − 84.6 × (syllables ÷ words). Higher reads easier. Syllables are estimated, and the formula was made for English.
- Most repeated
- library ×2
- quiet ×2
- Most frequent words, leaving out common ones like “the” and “and”. Repetition can be a choice; this only counts it.
None of these signals says who wrote a text. This experiment does not establish authorship.
Keeping the meaning
A meaning-preserving revision is an edit that still says the same thing. The drawing shows one way to check that, step by step.
Original
“The meeting, which had been planned for later in the week, was moved to Monday.”
Suggested edit
“The meeting moved from later in the week to Monday.”
Meaning check
Same event, same change, same day. Nothing added, nothing lost.
Keep or reject
Keep it. Had the day or the event changed, the original would stay.