word2vec (version 0.4.0)

word2vec_similarity: Similarity between word vectors as used in word2vec

Description

The similarity between word vectors is defined

• for type 'dot': as the square root of the average inner product of the vector elements (sqrt(sum(x . y) / ncol(x))) capped to zero

• for type 'cosine': as the the cosine similarity, namely sum(x . y) / (sum(x^2)*sum(y^2))

Usage

word2vec_similarity(x, y, top_n = +Inf, type = c("dot", "cosine"))

Value

By default, the function returns a similarity matrix between the rows of x and the rows of y. The similarity between row i of x and row j of y is found in cell [i, j] of the returned similarity matrix.

If top_n is provided, the return value is a data.frame with columns term1, term2, similarity and rank indicating the similarity between the provided terms in x and y

ordered from high to low similarity and keeping only the top_n most similar records.

Arguments

x

a matrix with embeddings where the rownames of the matrix provide the label of the term

y

a matrix with embeddings where the rownames of the matrix provide the label of the term

top_n

integer indicating to return only the top n most similar terms from y for each row of x. If top_n is supplied, a data.frame will be returned with only the highest similarities between x and y instead of all pairwise similarities

type

character string with the type of similarity. Either 'dot' or 'cosine'. Defaults to 'dot'.

See Also

word2vec

Examples

x <- matrix(rnorm(6), nrow = 2, ncol = 3)
rownames(x) <- c("word1", "word2")
y <- matrix(rnorm(15), nrow = 5, ncol = 3)
rownames(y) <- c("term1", "term2", "term3", "term4", "term5")

word2vec_similarity(x, y)
word2vec_similarity(x, y, top_n = 1)
word2vec_similarity(x, y, top_n = 2)
word2vec_similarity(x, y, top_n = +Inf)
word2vec_similarity(x, y, type = "cosine")
word2vec_similarity(x, y, top_n = 1, type = "cosine")
word2vec_similarity(x, y, top_n = 2, type = "cosine")
word2vec_similarity(x, y, top_n = +Inf, type = "cosine")

## Example with a word2vec model
path  <- system.file(package = "word2vec", "models", "example.bin")
model <- read.word2vec(path)
emb   <- as.matrix(model)

x <- emb[c("gastheer", "gastvrouw", "kamer"), ]
y <- emb
word2vec_similarity(x, x)
word2vec_similarity(x, y, top_n = 3)
predict(model, x, type = "nearest", top_n = 3)