# weighting_functions: Weighting Functions In svs: Tools for Semantic Vector Spaces

## Description

Local and global weighting functions.

## Usage

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21``` ```lw_tf(x) lw_raw(x) lw_log(x) lw_bin(x) gw_idf(x) gw_idf_alt(x) gw_gfidf(x) gw_nor(x) gw_ent(x) gw_bin(x) gw_raw(x) ```

## Arguments

 `x` A numeric matrix.

## Details

There are many local and global weighting functions. In this package, local weighting functions are prefixed with `lw_` and global weighting functions with `gw_`, so users can define their own weighting functions.

Local weighting functions (i.e. weighting every cell in the matrix):

• `lw_tf` Term frequency: f(x) = x.

• `lw_raw` Raw frequency, which is the same as the term frequency: f(x) = x.

• `lw_log` Logarithm: f(x) = log(x + 1).

• `lw_bin` Binary: f(x) = 1 if x > 0 and 0 otherwise.

Global weighting functions, weighting the columns of the matrix (hence, these weighting functions work according to expectation for a document-term matrix, i.e. with the documents as the rows and the terms as the columns):

• `gw_idf` Inverse document frequency: f(x) = log( nrow(x) / n + 1) where n = the number of rows in which the column >0.

• `gw_idf_alt` Alternative definition of the inverse document frequency: f(x) = log( nrow(x) / n) + 1 where n = the number of rows in which the column >0.

• `gw_gfidf` Global frequency multiplied by inverse document frequency: f(x) = colSums(x) / n where n = the number of rows in which the column >0.

• `gw_nor` Normal(ized) frequency: f(x) = x / colSums(x^2).

• `gw_ent` Entropy: f(x) = 1 + the relative Shannon entropy.

• `gw_bin` Binary: f(x) = 1.

• `gw_raw` Raw, which is the same as binary: f(x) = 1.

## Value

A numeric matrix.

`fast_lsa`.
 ```1 2 3 4 5``` ```SndT_Fra <- read.table(system.file("extdata", "SndT_Fra.txt", package = "svs"), header = TRUE, sep = "\t", quote = "\"", encoding = "UTF-8") tab.SndT_Fra <- table(SndT_Fra) lw_log(tab.SndT_Fra) gw_idf(tab.SndT_Fra) ```