Text Similarity Checker

How to compare two texts

Paste Text 1 and Text 2, choose any matching rules you need, then select Calculate similarity. Inspect the three scores or expand the token counts. The copied report includes the formulas, active rules and counts, but not your source text.

What the scores measure

Word tokens start with a Unicode letter or number and continue through letters, numbers or combining marks. Punctuation, including apostrophes, separates them; standalone marks are ignored. Unspaced text such as “你好世界” is one token here. No stopwords are removed, and word order does not affect the two word scores.

Edit similarity — Levenshtein

The distance is the fewest single-code-point insertions, deletions or replacements needed to turn one text into the other. Each costs 1; swapping two different adjacent code points takes 2 edits. The score is 1 − distance ÷ the longer code-point length. If both texts are empty after matching rules, the score is 100% by convention. Code points are not displayed characters: an emoji or accented character can contain several.

Word overlap — Jaccard

Shared distinct tokens ÷ all distinct tokens found across both texts. Repetitions do not change the score. It is unavailable when neither text has any word tokens.

Word frequency — cosine

Each token’s frequency forms a coordinate in a count vector. The score is the dot product divided by the product of the two vector lengths. Repetitions matter. It is unavailable when either text has no word tokens. This calculation uses plain counts, without TF-IDF weights or a language model.

Reading the result

Similar spelling or shared vocabulary does not establish the same meaning. “The door is open” and “The door is not open” share most of their words, but make different assertions. This tool compares only the texts you supply; it does not search the web, detect plagiarism or judge authorship. Values close to 0% or 100% are displayed as bounds so rounding does not turn a small difference into an exact match.