CorpusData: Manage Corpus Data and Encode CWB Corpus.

CorpusDataR Documentation

Manage Corpus Data and Encode CWB Corpus.

Description

Manage Corpus Data and Encode CWB Corpus.

Manage Corpus Data and Encode CWB Corpus.

Details

See the CWB Encoding Tutorial on characters allowed for encoding attributes: "By convention, all attribute names must be lowercase (more precisely, they may only contain the characters a-z, 0-9, -, and _, and may not start with a digit). Therefore, the names of XML elements to be included in the CWB corpus must not contain any non-ASCII or uppercase letters." (section 2)

Import XML files.

Public fields

chunktable

A data.table with column "id" (unique values), columns with metadata, and a column with text chunks.

tokenstream

A data.table with a column "cpos" (corpus position), and columns with positional attributes, such as "word", "lemma", "pos", "stem".

metadata

A data.table with a column "id", to link data with chunks/tokenstream, columns with document-level metadata, and a column "cpos_left" and "cpos_right", which can be generated using method ⁠$add_corpus_positions()⁠.

sentences

A data.table.

named_entities

A data.table.

Methods

Public methods


Method new()

Initialize a new instance of class CorpusData.

Usage
CorpusData$new()
Returns

A class CorpusData object.


Method print()

Print summary of CorpusData object.

Usage
CorpusData$print()

Method tokenize()

Simple tokenization of text in chunktable.

Usage
CorpusData$tokenize(..., verbose = TRUE, progress = TRUE)
Arguments
...

Arguments that are passed into tokenizers::tokenize_words().

verbose

A logical value, whether to be verbose.

progress

A logical value, whether to show progress bar.


Method import_xml()

Usage
CorpusData$import_xml(
  filenames,
  body = "//body",
  meta = NULL,
  mc = NULL,
  progress = TRUE
)
Arguments
filenames

A vector of files to process.

body

An xpath expression defining the body of the XML document.

meta

A named character vector with XPath expressions.

mc

A numeric/integer value, number of cores to use.

progress

A logical value, whether to show progress bar.

Returns

The CorpusData object is returned invisibly.


Method add_corpus_positions()

Add column 'cpos' to tokenstream and columns 'cpos_left' and 'cpos_right' to metadata.

Usage
CorpusData$add_corpus_positions(verbose = TRUE)
Arguments
verbose

A logical value, whether to be verbose.


Method purge()

Remove patterns from chunkdata that are known to cause problems. This is done most efficiently at the chunkdata level of data preparation as the length of the character vector to handle is much smaller than when tokenization/annotation has been performed.

Usage
CorpusData$purge(
  replacements = list(c("^\\s*<.*?>\\s*$", ""), c("’", "'"))
)
Arguments
replacements

A list of length-two character vectors with regular expressions and replacements.


Method encode()

Encode corpus. If the corpus already exists, it will be removed.

Usage
CorpusData$encode(
  corpus,
  p_attributes = "word",
  s_attributes = NULL,
  encoding,
  registry_dir = Sys.getenv("CORPUS_REGISTRY"),
  data_dir = NULL,
  method = c("R", "CWB"),
  verbose = TRUE,
  compress = FALSE,
  reload = TRUE,
  quietly = TRUE
)
Arguments
corpus

The name of the CWB corpus.

p_attributes

Positional attributes.

s_attributes

Columns that will be encoded as structural attributes.

encoding

Encoding/charset of the CWB corpus.

registry_dir

Corpus registry, the directory where registry files are stored.

data_dir

Directory where to create directory for indexed corpus files.

method

Either "R" or "CWB".

verbose

A logical value, whether to be verbose.

compress

A logical value, whether to compress corpus.

reload

A logical value, whether to reload corpus.

quietly

A logical value passed into RcppCWB::cwb_makeall(), RcppCWB::cwb_huffcode() and RcppCWB::cwb_compress_rdx to control verbosity of these functions.


Method clone()

The objects of this class are cloneable with this method.

Usage
CorpusData$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Examples

library(RcppCWB)
library(data.table)

# this example relies on the R method to write data to disk, there is also a method "CWB"
# that relies on CWB tools to generate the indexed corpus. The CWB can downloaded
# and installed within the package by calling cwb_install()

# create temporary registry file so that data in RcppCWB package can be used

registry_rcppcwb <- system.file(package = "RcppCWB", "extdata", "cwb", "registry")
registry_tmp <- fs::path(tempdir(), "registry")
if (!dir.exists(registry_tmp)) dir.create(registry_tmp)
r <- registry_file_parse("REUTERS", registry_dir = registry_rcppcwb)
r[["home"]] <- system.file(package = "RcppCWB", "extdata", "cwb", "indexed_corpora", "reuters")
registry_file_write(r, corpus = "REUTERS", registry_dir = registry_tmp)

# decode structural attribute 'places'

s_attrs_places <- RcppCWB::s_attribute_decode(
  corpus = "REUTERS",
  data_dir = system.file(package = "RcppCWB", "extdata", "cwb", "indexed_corpora", "reuters"),
  s_attribute = "places", method = "R"
)
s_attrs_places[["id"]] <- 1L:nrow(s_attrs_places)
setnames(s_attrs_places, old = "value", new = "places")

# decode positional attribute 'word'

tokens <- apply(s_attrs_places, 1, function(row){
  ids <- cl_cpos2id(
    corpus = "REUTERS", cpos = row[1]:row[2],
    p_attribute = "word", registry = registry_tmp
  )
  cl_id2str(corpus = "REUTERS", id = ids, p_attribute = "word", registry = registry_tmp)
})
tokenstream <- rbindlist(
lapply(
  1L:length(tokens),
  function(i) data.table(id = i, word = tokens[[i]]))
  )
tokenstream[["cpos"]] <- 0L:(nrow(tokenstream) - 1L)

# create CorpusData object (see vignette for further explanation)

CD <- CorpusData$new()
CD$tokenstream <- as.data.table(tokenstream)
CD$metadata <- as.data.table(s_attrs_places)

# Remove temporary registry with home dir still pointing to RcppCWB data dir
# to prevent data from being deleted
file.remove(fs::path(registry_tmp, "reuters"))
file.remove(registry_tmp)

# create temporary directories (registry directory and one for indexed corpora)

registry_tmp <- fs::path(tempdir(), "registry")
data_dir_tmp <- fs::path(tempdir(), "data_dir")
if (!dir.exists(registry_tmp)) dir.create(registry_tmp)
if (!dir.exists(data_dir_tmp)) dir.create(data_dir_tmp)

CD$encode(
  corpus = "REUTERS", encoding = "utf8",
  p_attributes = "word", s_attributes = "places",
  registry_dir = registry_tmp, data_dir = data_dir_tmp,
  method = "R"
)
reg <- registry_data(name = "REUTERS", id = "REUTERS", home = data_dir_tmp, p_attributes = "word")
registry_file_write(data = reg, corpus = "REUTERS", registry_dir = registry_tmp)

# see whether it works

cl_cpos2id(corpus = "REUTERS", p_attribute = "word", cpos = 0L:4049L, registry = registry_tmp)

PolMine/cwbtools documentation built on March 5, 2024, 10:21 a.m.