MinHash Jaccard Similarity Estimator

Compare two sets of strings using seeded 32-bit MinHash signatures. Estimated Jaccard can differ from exact Jaccard, especially with few hashes.

Description

Compare two sets of strings using seeded 32-bit MinHash signatures. Estimated Jaccard can differ from exact Jaccard, especially with few hashes.

MinHash Jaccard Similarity Estimator: Compare two sets of strings using seeded 32-bit MinHash signatures. Estimated Jaccard can differ from exact Jaccard, especially with few hashes.

When to use MinHash Jaccard Similarity

Use this probabilistic structure when bounded memory and fast approximate membership, cardinality, or similarity queries are preferable to retaining every original item.

left: Required list input.
right: Required list input.
hash Count: Required integer input.

How MinHash Jaccard Similarity works

Compare two sets of strings using seeded 32-bit MinHash signatures. Estimated Jaccard can differ from exact Jaccard, especially with few hashes. The tool evaluates the supplied inputs together and returns the named outputs below; it does not infer omitted operating conditions or change the units shown.^[1]

estimated Jaccard: The resulting estimated jaccard returned as a number.
exact Jaccard: The resulting exact jaccard returned as a number.
left Signature: The resulting left signature returned as a list.
right Signature: The resulting right signature returned as a list.

Worked example

For example, left = [“a”, “b”], right = [“b”, “c”], hash Count = 32 produces estimated Jaccard = 0.375, exact Jaccard = 0.3333333333333333, left Signature = [1185767313, 1185767313, 444641715, 2111704691, 1377171301, 1377171301, 2111704691, 1311273838, 1358841603, 1358841603, 1311273838, 250914211, 2249599777, 2249599777, 250914211, 590027118, 550558746, 550558746, 590027118, 910896560, 1701362178, 1701362178, 910896560, 228143771, 794621484, 794621484, 228143771, 2437239260, 169220629, 169220629, 2437239260, 1294499476], right Signature = [1185767313, 444641715, 444641715, 1377171301, 1377171301, 2111704691, 2111704691, 2144823071, 2144823071, 1311273838, 1311273838, 1911777871, 1911777871, 250914211, 250914211, 550558746, 550558746, 1560842042, 1560842042, 1701362178, 1701362178, 910896560, 910896560, 228143771, 228143771, 794621484, 794621484, 2693131916, 2693131916, 169220629, 169220629, 2416073818].

Limitations and assumptions

Probabilistic structures trade exactness for space. Error rates depend on capacity, hash quality, parameter choice, and merge compatibility; some structures permit false positives, and most cannot reconstruct the inserted values.
Use finite inputs in the displayed units and preserve more precision than the final presentation requires. Independently verify safety-critical, financial, compliance, or production decisions.

Alternative or Complementary approaches

Choose parameters from the expected cardinality and acceptable error rate, monitor saturation, and retain an exact store when results must be verified. Merge only structures created with compatible parameters and hashing.

References

MinHash — Wikipedia contributors

Similar or alternative tools

Locality-Sensitive Hash Candidate Checker
Use banded MinHash signatures to flag a pair of string sets as a similarity candidate. A matching band is not proof of semantic similarity.
HyperLogLog Cardinality Estimator
Estimate the number of distinct strings with a fixed-size HyperLogLog sketch. The estimate is probabilistic and may differ from the exact count.
Counting Bloom Filter Calculator
Build a deterministic counting Bloom filter, apply requested removals, and test a query. Positive membership and multiplicity remain probabilistic because hash collisions can overestimate both.