1) Tokenize. Your text is split into words, numbers, and punctuation. Non-ASCII is stripped so Blogger-safe output stays clean.
2) Guess role. Each word gets a rough part-of-speech guess (noun, verb, adj, adv, number, proper-noun) using simple suffix and case cues.
3) Infer topic. Each token is mapped to a topic field (law, money, body, etc.). With Bias set to Auto, the word itself nudges the choice.
4) Decide to swap. Replacement probability comes from Style Intensity plus a small sentiment tilt from Mood.
5) Substitute. If swapping, the word is replaced from the topic lexicon by role (noun, verb, adj, adv). Case is preserved.
6) Normalize. Spacing and punctuation are cleaned, then lines wrap to your Line Width.
Seeds. Enter a seed or click Reroll. Same seed with same options reproduces output.
Privacy. All logic runs in your browser only.
Shortcuts: ? open, Esc close.
No comments:
Post a Comment