Dicekerroin

Dicekerroin, also called the Dice coefficient, is a statistical measure of similarity between two samples. In Finnish contexts it is known as Dicekerroin and is frequently referred to as the Sørensen–Dice coefficient, named after Sørensen and Dice. For two finite sets A and B, the coefficient is defined as Dice(A,B) = 2|A ∩ B| / (|A| + |B|). When applied to binary vectors, it can be written as 2TP / (2TP + FP + FN). The value ranges from 0 to 1, with 1 indicating identical content and 0 indicating no overlap.

The Dice coefficient is widely used in natural language processing, information retrieval and data deduplication to measure similarity between token sets, feature sets, or presence/absence data. In text applications, A and B might be the sets of words or n-grams in two documents; in bioinformatics, they may represent the presence of features or genes.

Extensions and related concepts include generalized or weighted Dice coefficients for non-binary data, and adaptations for multi-label or imbalanced data. The Dice coefficient is related to the Jaccard index, with Dice = 2J / (1+J).

Limitations include sensitivity to small set sizes and to how the data are binarized or tokenized; it emphasizes overlap but ignores non-overlapping elements, which can be problematic when sets differ greatly in size. Despite these caveats, the Dice coefficient remains a widely used, simple measure of set similarity across disciplines.