Jaccard Coefficient

What:

An early way of measuring how well a query matches a document in Information Retrieval (IR). This was then used for ranked information retrieval.

How:

• $J(A,B)=\frac{|A\cap B|}{|A\cup B|}$ (Intersection over union)

Limitations:

• Ignores term frequency and term rarity.
  • Common words like “the” is equal to a rare, important word.