String Similarity Checker

How to use

1. 1Paste one string into A and another into B. Try one of the sample pairs to see how typos and reworded text score across metrics.
2. 2Set preprocessing: case insensitive, trim or collapse whitespace, and Unicode NFKC normalization. These are applied before any metric is calculated.
3. 3Pick the n-gram size for Jaccard, Cosine, and Dice-n (bigrams are a sensible default; trigrams work well for longer text).
4. 4Read the Overall similarity badge for a quick summary, then check the individual metric cards: each has a meter, a percent score, and a one-line explanation of when to use it.
5. 5Use Swap A and B if you want to verify symmetry, then Copy report to grab a clean text summary of every metric for your notes, ticket, or dataset documentation.

About this tool

String Similarity Checker compares two strings with every common similarity and distance metric at the same time, so you can pick the right one for your job without bouncing between calculators. Paste a string into A and another into B and the tool reports the Levenshtein edit distance and its normalized similarity, the Damerau-Levenshtein distance (which counts an adjacent-letter transposition as a single edit and is better for typo correction), the Hamming distance (defined only when the two strings have the same length, useful for fixed-length codes, hashed fingerprints, and DNA k-mers), the longest common subsequence length and the LCS ratio used by diff tools, the Jaro similarity, the Jaro-Winkler similarity (Jaro plus a prefix-match bonus, originally built at the US Census Bureau for matching personal names), the Dice / Sorensen coefficient on character bigrams, the Dice coefficient on n-grams of your chosen size, the Jaccard index on n-gram sets (intersection over union), and the cosine similarity on n-gram count vectors (the standard for vector-space retrieval and many NLP pipelines). A single composite Overall similarity score blends the four most-cited metrics for a quick at-a-glance read. Preprocessing options apply before any metric runs, so you can compare strings while ignoring case, trimming or collapsing whitespace, or applying Unicode NFKC normalization (which makes the 'fi' ligature equal to 'fi' and folds full-width characters). The n-gram size for Jaccard, Cosine, and Dice-n is configurable between 1 and 8 so you can tune for short names (bigrams) or longer text fields (trigrams or 4-grams). Every algorithm is implemented from scratch in TypeScript and runs in your browser; there is no library dependency to audit and no server round trip. Useful for record-linkage and fuzzy dedupe pipelines, spell-checkers and typo-tolerant search, name and address matching, plagiarism and paraphrase checks, fuzzy SKU and EAN matching, manual review of database join keys, building or sanity-checking a similarity threshold, learning what each metric actually measures, and homework or interview prep on edit-distance algorithms. Input is capped at 5,000 characters per side, which keeps the O(n*m) algorithms responsive on a laptop while still covering paragraph-length text and any realistic dedup key. Nothing is uploaded.

Free to use. Works in your browser. No signup, no login.