quanteda.textstats (version 0.97.2)

textstat_frequency: Tabulate feature frequencies

Description

Produces counts and document frequencies summaries of the features in a dfm, optionally grouped by a docvars variable or other supplied grouping variable.

Usage

textstat_frequency(
  x,
  n = NULL,
  groups = NULL,
  ties_method = c("min", "average", "first", "random", "max", "dense"),
  ...
)

Value

a data.frame containing the following variables:

feature

(character) the feature

frequency

count of the feature

rank

rank of the feature, where 1 indicates the greatest frequency

docfreq

document frequency of the feature, as a count (the number of documents in which this feature occurred at least once)

docfreq

document frequency of the feature, as a count

group

(only if groups is specified) the label of the group. If the features have been grouped, then all counts, ranks, and document frequencies are within group. If groups is not specified, the group column is omitted from the returned data.frame.

textstat_frequency returns a data.frame of features and their term and document frequencies within groups.

Arguments

x

a dfm object

n

(optional) integer specifying the top n features to be returned, within group if groups is specified

groups

grouping variable for sampling, equal in length to the number of documents. This will be evaluated in the docvars data.frame, so that docvars may be referred to by name without quoting. This also changes previous behaviours for groups. See news(Version >= "3.0", package = "quanteda") for details.

ties_method

character string specifying how ties are treated. See base::rank() for details. Unlike that function, however, the default is "min", so that frequencies of 10, 10, 11 would be ranked 1, 1, 3.

...

additional arguments passed to dfm_group(). This can be useful in passing force = TRUE, for instance, if you are grouping a dfm that has been weighted.

Examples

library("quanteda")
set.seed(20)
dfmat1 <- dfm(tokens(c("a a b b c d", "a d d d", "a a a")))

textstat_frequency(dfmat1)
textstat_frequency(dfmat1, groups = c("one", "two", "one"), ties_method = "first")
textstat_frequency(dfmat1, groups = c("one", "two", "one"), ties_method = "average")

dfmat2 <- corpus_subset(data_corpus_inaugural, President == "Obama") %>%
   tokens(remove_punct = TRUE) %>%
   tokens_remove(stopwords("en")) %>%
   dfm()
tstat1 <- textstat_frequency(dfmat2)
head(tstat1, 10)

dfmat3 <- head(data_corpus_inaugural) %>%
   tokens(remove_punct = TRUE) %>%
   tokens_remove(stopwords("en")) %>%
   dfm()
textstat_frequency(dfmat3, n = 2, groups = President)


if (FALSE) {
# plot 20 most frequent words
library("ggplot2")
ggplot(tstat1[1:20, ], aes(x = reorder(feature, frequency), y = frequency)) +
    geom_point() +
    coord_flip() +
    labs(x = NULL, y = "Frequency")

# plot relative frequencies by group
dfmat3 <- data_corpus_inaugural %>%
    corpus_subset(Year > 2000) %>%
    tokens(remove_punct = TRUE) %>%
    tokens_remove(stopwords("en")) %>%
    dfm() %>%
    dfm_group(groups = President) %>%
    dfm_weight(scheme = "prop")

# calculate relative frequency by president
tstat2 <- textstat_frequency(dfmat3, n = 10, groups = President)

# plot frequencies
ggplot(data = tstat2, aes(x = factor(nrow(tstat2):1), y = frequency)) +
    geom_point() +
    facet_wrap(~ group, scales = "free") +
    coord_flip() +
    scale_x_discrete(breaks = nrow(tstat2):1,
                       labels = tstat2$feature) +
    labs(x = NULL, y = "Relative frequency")
}