polmineR (version 0.7.11)

cooccurrences: Get cooccurrence statistics.

Description

Get cooccurrence statistics.

Usage

cooccurrences(.Object, ...)
# S4 method for character
cooccurrences(.Object, query, cqp = is.cqp,
  p_attribute = getOption("polmineR.p_attribute"), s_attribute = NULL,
  left = getOption("polmineR.left"),
  right = getOption("polmineR.right"), stoplist = NULL,
  positivelist = NULL, regex = FALSE, keep = NULL, cpos = NULL,
  method = "ll", mc = getOption("polmineR.mc"), verbose = FALSE,
  progress = FALSE, ...)
# S4 method for partition
cooccurrences(.Object, query, cqp = is.cqp,
  left = getOption("polmineR.left"),
  right = getOption("polmineR.right"),
  p_attribute = getOption("polmineR.p_attribute"), s_attribute = NULL,
  stoplist = NULL, positivelist = NULL, keep = NULL, method = "ll",
  mc = FALSE, progress = TRUE, verbose = FALSE, ...)
# S4 method for context
cooccurrences(.Object, method = "ll",
  verbose = FALSE)
# S4 method for Corpus
cooccurrences(.Object, query,
  p_attribute = getOption("polmineR.p_attribute"), ...)
# S4 method for partition_bundle
cooccurrences(.Object, query,
  mc = getOption("polmineR.mc"), ...)
# S4 method for Cooccurrences
cooccurrences(.Object, query)

Arguments

.Object

A partition object, or a character vector with a CWB corpus.

...

Further parameters that will be passed into bigmatrix (applies only of big = TRUE).

query

A query, either a character vector to match a token, or a CQP query.

cqp

Defaults to is.cqp-function, or provide TRUE/FALSE; relevant only if query is not NULL.

p_attribute

The p-attribute of the tokens/the query.

s_attribute

If provided, it will be checked that corpus positions of windows do not extend beyond the region defined by the s-attribute.

left

Number of tokens to the left of the query match.

right

Number of tokens to the right of the query match.

stoplist

Exclude a query hit from analysis if stopword(s) is/are in context (relevant only if query is not NULL).

positivelist

Character vector or numeric vector: include a query hit only if token in positivelist is present. If positivelist is a character vector, it is assumed to provide regex expressions (incredibly long if the list is long) (relevant only if query is nut NULL)

regex

A logical value, whether stoplist/positivelist are interpreted as regular expressions.

keep

list with tokens to keep

cpos

integer vector with corpus positions, defaults to NULL - then the corpus positions for the whole corpus will be used

method

The statistical test(s) to use (defaults to "ll").

mc

whether to use multicore

verbose

A logical value, whether to be verbose.

progress

A logical value, whether to output progress bar.

Value

a cooccurrences-class object

References

Baker, Paul (2006): Using Corpora in Discourse Analysis. London: continuum, p. 95-120 (ch. 5).

Manning, Christopher D.; Schuetze, Hinrich (1999): Foundations of Statistical Natural Language Processing. MIT Press: Cambridge, Mass., pp. 151-189 (ch. 5).

See Also

See the documentation for the ll-method for an explanation of the computation of the log-likelihood statistic.

Examples

# NOT RUN {
use("polmineR")
merkel <- partition("GERMAPARLMINI", interjection = "speech", speaker = ".*Merkel", regex = TRUE)
merkel <- enrich(merkel, p_attribute = "word")
cooc <- cooccurrences(merkel, query = "Deutschland")

# use subset-method to filter results
a <- cooccurrences("REUTERS", query = "oil")
b <- subset(a, !is.na(ll))
c <- subset(b, !word %in% tm::stopwords("en"))
d <- subset(c, count_coi >= 5)
e <- subset(c, ll >= 10.83)
format(e)

# using pipe operator may be convenient
if (require(magrittr)){
cooccurrences("REUTERS", query = "oil") %>%
  subset(!is.na(ll)) %>%
  subset(!word %in% tm::stopwords("en")) %>%
  subset(count_coi >= 5) %>%
  subset(ll >= 10.83) %>%
  format()
}
# }