Compute Cosine Similarity

Cosine similarity between embedding rows, for example to check how close two texts are in meaning before or after topic modeling. Rows are L2-normalized first, so unnormalized embeddings give the same result.

Usage

topic_similarity(x, y = NULL)

Arguments

x

Numeric embedding matrix, or one numeric vector.

y

Optional second embedding matrix or vector. When omitted, computes all pairwise similarities among rows of x.

Value

A numeric matrix of cosine similarities in [-1, 1] with one row per row of x and one column per row of y (or of x when y is omitted). A vector counts as a one-row matrix.

Examples

x <- rbind(c(1, 0), c(0, 1), c(1, 1))
topic_similarity(x)
#>           [,1]      [,2]      [,3]
#> [1,] 1.0000000 0.0000000 0.7071068
#> [2,] 0.0000000 1.0000000 0.7071068
#> [3,] 0.7071068 0.7071068 1.0000000