1001Ferramentas
📈Text

Jaccard Similarity Calculator

Calculate the Jaccard similarity index between two texts by comparing their shared sets of words. Great for plagiarism detection and measuring overlap.

Jaccard

How much two texts share their vocabulary

Two exam answers that look too alike, two versions of the same paragraph, a thousand product titles with duplicates: the question is always how much vocabulary they have in common. The Jaccard index answers with a number between 0 and 1, the count of shared words divided by the count of distinct words across both texts together.

Each text becomes a set of unique lowercase words, so order and repetition disappear. O gato mordeu o cachorro and o cachorro mordeu o gato score 1, a perfect match, even though they say opposite things. The sample in the box gives 0.333333, two shared words over six distinct ones. Word splitting treats only unaccented letters and digits as word characters, so coração breaks into cora and o.

With no stopword list, long Portuguese texts always share de, a, o, que and para, which pushes the index up. For plagiarism work, strip the function words before pasting or compare short passages, and keep in mind that a paraphrase built on synonyms scores low while still being copied. Very small values print in scientific notation. Both texts stay in your browser.

Frequently asked questions

Is 0.3 high or low?
It depends on length and language. In Portuguese, two unrelated texts already reach 0.1 or 0.2 on prepositions alone. Compare similar pairs against each other instead of reading one value on its own.
Does word order matter?
No, it is a set comparison. To take order into account you need n-gram similarity or an edit distance.
Do accents affect the score?
Somewhat. The splitter treats accented letters as punctuation, so coração breaks into pieces. Both texts split the same way, so comparisons hold, but the absolute value shifts.

Related Tools