tm (version 0.6-2)

# Zipf_n_Heaps: Explore Corpus Term Frequency Characteristics

## Description

Explore Zipf's law and Heaps' law, two empirical laws in linguistics describing commonly observed characteristics of term frequency distributions in corpora.

## Usage

Zipf_plot(x, type = "l", ...)
Heaps_plot(x, type = "l", ...)

## Arguments

x
a document-term matrix or term-document matrix with unweighted term frequencies.
type
a character string indicating the type of plot to be drawn, see plot.
...
further graphical parameters to be used for plotting.

## Value

The coefficients of the fitted linear model. As a side effect, the corresponding plot is produced.

## Details

Zipf's law (e.g., http://en.wikipedia.org/wiki/Zipf%27s_law) states that given some corpus of natural language utterances, the frequency of any word is inversely proportional to its rank in the frequency table, or, more generally, that the pmf of the term frequencies is of the form $c k^{-\beta}$, where $k$ is the rank of the term (taken from the most to the least frequent one). We can conveniently explore the degree to which the law holds by plotting the logarithm of the frequency against the logarithm of the rank, and inspecting the goodness of fit of a linear model.

Heaps' law (e.g., http://en.wikipedia.org/wiki/Heaps%27_law) states that the vocabulary size $V$ (i.e., the number of different terms employed) grows polynomially with the text size $T$ (the total number of terms in the texts), so that $V = c T^\beta$. We can conveniently explore the degree to which the law holds by plotting $\log(V)$ against $\log(T)$, and inspecting the goodness of fit of a linear model.

## Examples

data("acq")
m <- DocumentTermMatrix(acq)
Zipf_plot(m)
Heaps_plot(m)