analyzeText

Analyze a text string and return heuristic scores and text counts.

Syntax

const result = WriteLikeMe.analyzeText(text);

Parameters

Parameter	Type	Description
text	string	Text to analyze. Non-string values return the empty-input result.

Returns

Property	Type	Current behavior
formality	number (0-1)	Keyword ratio from a small set of formal and casual words
complexity	number (0-1)	Average word length and proportion of words with seven or more characters
sentence_length	number (0-1)	Average words per sentence, normalized against 30 words
vocabulary	number (0-1)	Unique-token count adjusted by the square root of total token count
tone	number (0-1)	Ratio of words in small positive/negative keyword lists; 0.5 if neither appears
creativity	number (0-1)	Weighted combination of vocabulary variety, imagery keywords, and punctuation variety
punctuation	number (0-1)	Share of commas among commas and periods; 0.5 when neither appears
confidence	number (0-1)	Word count divided by 200, capped at 1; not statistical confidence
wordCount	number	Count of Unicode letter/number tokens
charCount	number	UTF-16 code-unit count after trimming leading and trailing whitespace
commaCount	number	Number of comma characters
periodCount	number	Number of period characters
commaShare	number (0-1)	Same comma-to-period balance as punctuation
semicolonCount, colonCount, dashCount	number	Counts of these punctuation marks
questionCount, exclamationCount	number	Question and exclamation mark counts
first_person, second_person	number (0-1)	Scaled rates of first- and second-person pronouns
contraction_rate	number (0-1)	Scaled contraction frequency
sentence_variation	number (0-1)	Coefficient of variation for sentence lengths, capped at 1
paragraph_length	number (0-1)	Average paragraph word count normalized against 60 words
question_rate, exclamation_rate	number (0-1)	Mark counts per sentence, capped at 1
transition_rate, hedging_rate	number (0-1)	Scaled frequency of words from small English-language lists
frequentWords	string[]	Up to five most frequent content words, excluding a small common-word list

Non-string, empty, and whitespace-only input returns the default category values, zero confidence, zero word/character/punctuation counts, zero voice-cue scores, an empty frequentWords list, and neutral punctuation balance.

The punctuation signal measures comma-versus-period habits, not grammatical correctness. Its SVG band uses comma-shaped marks and period dots to make that balance visually distinct.

Example

const WriteLikeMe = require('@ai-clyde/writelikeme');

const result = WriteLikeMe.analyzeText('Hello, world!');
console.log(result.wordCount); // 2
console.log(result.charCount); // 13
console.log(result.complexity); // 0.28
console.log(result.vocabulary); // 0.57
console.log(result.confidence); // 0.01

Tokenization recognizes Unicode letters and numbers. The keyword lists are small, English-focused, and do not understand context, negation, sarcasm, or word meaning outside the listed terms. The sentence and word-length metrics are simple proxies. Results are experimental, may be inaccurate (especially for short text or other languages), and must not be treated as a reliable linguistic or psychological measurement or used for consequential decisions.

Compare a draft with a reference

Use compareTextStyle(referenceText, candidateText) to check whether an AI draft resembles the user's analyzed style:

const comparison = WriteLikeMe.compareTextStyle(userSample, generatedDraft);
console.log(comparison.similarity); // 0-1 heuristic similarity
console.log(comparison.differences.formality); // Absolute difference for a style signal
console.log(comparison.differences.frequentWordsOverlap); // 0-1 word overlap

similarity is the average similarity across the seven SVG categories, numeric voice cues, and frequent-word overlap, rounded to two decimal places. differences contains the absolute difference for each numeric signal and frequentWordsOverlap, the Jaccard overlap of the two frequentWords lists (higher is a closer match). It compares style cues, not meaning, and cannot guarantee that a draft sounds like its author. If either text has no analyzable words, similarity is 0.

The SVG's visible pattern uses the seven categories accepted by generateStyleSVG. Its embedded metadata also stores the additional voice-cue scores and frequent words; these are not separate SVG bands. Use readStyleCode(svg) to retrieve that profile.