A high-level function that controls a number of other functions responsible
for dealing with a raw corpus stored as list, including deleting markup,
sampling from texts, converting samples to n-grams, etc. It is build on top
of a number of functions and thus it requires a large number of arguments.
The only obligatory argument, however, is an R object containing a raw corpus:
it is either an object of the class sylo.corpus, or a list of vectors,
their elements being particular texts.
parse.corpus(input.data, markup.type = "plain",
corpus.lang = "English", splitting.rule = NULL,
sample.size = 10000, sampling = "no.sampling",
sample.overlap = 0, number.of.samples = 1,
sampling.with.replacement = FALSE, features = "w",
ngram.size = 1, preserve.case = FALSE,
encoding = "UTF-8")The function returns an object of the class stylo.corpus. It is a list
containing as elements the samples (entire texts or sampled subsets) split into
words/characters and combined into n-grams (if applicable).
a list (preferably of the class stylo.corpus)
containing a raw corpus, i.e. a vector of texts.
choose one of the following values: plain
(nothing will happen), html (all tags will be deleted as well
as HTML header), xml (TEI header, any text between <note> </note>
tags, and all the tags will be deleted), xml.drama (as above;
additionally, speaker's names will be deleted, or strings within the
<speaker> </speaker> tags), xml.notitles (as above; but,
additionally, all the chapter/section (sub)titles will be deleted,
or strings within each the <head> </head> tags);
see delete.markup for further details.
an optional argument indicating the language of the texts
analyzed; the values that will affect the function's behavior are:
English.contr, English.all, Latin.corr (type
help(txt.to.words.ext) for explanation). The default value
is English.
if you are not satisfied with the default language
settings (or your input string of characters is not a regular text,
but a sequence of, say, dance movements represented using symbolic signs),
you can indicate your custom splitting regular expression here. This
option will overwrite the above language settings. For further details,
refer to help(txt.to.words).
desired size of samples, expressed in number of words; default value is 10,000.
one of three values: no.sampling (default),
normal.sampling, random.sampling. See make.samples
for explanation.
if this opion is used, a reference text is segmented
into consecutive, equal-sized samples that are allowed to partially
overlap. If one specifies the sample.size parameter of 5,000 and
the sample.overlap of 1,000, for example, the first sample of a text
contains words 1--5,000, the second 4001--9,000, the third sample 8001--13,000,
and so forth.
optional argument which will be used only if
random.sampling was chosen; it is self-evident.
optional argument which will be used only
if random.sampling was chosen; it specifies the method used to
randomly harvest words from texts.
an option for specifying the desired type of features:
w for words, c for characters (default: w). See
txt.to.features for further details.
an optional argument (integer) specifying the value of n,
or the size of n-grams to be produced. If this argument is missing,
the default value of 1 is used. See txt.to.features for further
details.
whether ot not to lowercase all characters in the corpus (default = F).
useful if you use Windows and non-ASCII alphabets: French,
Polish, Hebrew, etc. In such a situation, it is quite convenient to
convert your text files into Unicode and to set this option to
encoding = "UTF-8". In Linux and Mac, you are always expected
to use Unicode, thus you don't need to set anything.
Maciej Eder
load.corpus.and.parse, delete.markup,
txt.to.words, txt.to.words.ext,
txt.to.features, make.samples
if (FALSE) {
data(novels)
# depending on the size of the corpus, it might take a while:
parse.corpus(novels)
}
Run the code above in your browser using DataLab