koRpus (version 0.06-5)

treetag: A function to call TreeTagger

Description

This function calls a local installation of TreeTagger[1] to tokenize and POS tag the given text.

Usage

treetag(file, treetagger = "kRp.env", rm.sgml = TRUE, lang = "kRp.env", apply.sentc.end = TRUE, sentc.end = c(".", "!", "?", ";", ":"), encoding = NULL, TT.options = NULL, debug = FALSE, TT.tknz = TRUE, format = "file", stopwords = NULL, stemmer = NULL)

Arguments

file

Either a connection or a character vector, valid path to a file, containing the text to be analyzed. If file is a connection, its contents will be written to a temporary file, since TreeTagger can't read from R connection objects.

treetagger

A character vector giving the TreeTagger script to be called. If set to "kRp.env" this is got from get.kRp.env. Only if set to "manual", it is assumend not to be a wrapper script that can work the given text file, but that you would like to manually tweak options for tokenizing and POS tagging yourself. In that case, you need to provide a full set of options with the TT.options parameter.

rm.sgml

Logical, whether SGML tags should be ignored and removed from output

lang

A character string naming the language of the analyzed corpus. See kRp.POS.tags for all supported languages. If set to "kRp.env" this is got from get.kRp.env.

apply.sentc.end

Logical, whethter the tokens defined in sentc.end should be searched and set to a sentence ending tag.

sentc.end

A character vector with tokens indicating a sentence ending. This adds to TreeTaggers results, it doesn't really replace them.

encoding

A character string defining the character encoding of the input file, like "Latin1" or "UTF-8". If NULL, the encoding will either be taken from a preset (if defined in TT.options), or fall back to "". Hence you can overwrite the preset encoding with this parameter.

TT.options

A list of options to configure how TreeTagger is called. You have two basic choices: Either you choose one of the pre-defined presets or you give a full set of valid options:

• path Mandatory: The absolute path to the TreeTagger root directory. That is where its subfolders bin, cmd and lib are located.
• preset Optional: If you choose one of the pre-defined presets here:
  • "de-utf8" German, UTF-8
  • "de" German
  • "en" English
  • "es-utf8" Spanish, UTF-8
  • "es" Spanish
  • "fr-utf8" French, UTF-8
  • "fr" French
  • "it-utf8" Italian, UTF-8
  • "it" Italian
  • "ru" Russian, UTF-8

  you can omit all the following elements, because they will be filled with defaults. Of course this only makes sense if you have a working default installation.

• tokenizer Mandatory: A character string, naming the tokenizer to be called. Interpreted relative to path/cmd/.
• tknz.opts Optional: A character string with the options to hand over to the tokenizer. You don't need to specify "-a" if abbrev is given. If TT.tknz=FALSE, you can pass configurational options to tokenize by provinding them as a named list (instead of a character string) here.
• tagger Mandatory: A character string, naming the tagger-command to be called. Interpreted relative to path/bin/.
• abbrev Optional: A character string, naming the abbreviation list to be used. Interpreted relative to path/lib/.
• params Mandatory: A character string, naming the parameter file to be used. Interpreted relative to path/lib/.
• lexicon Optional: A character string, naming the lexicon file to be used. Interpreted relative to path/lib/.
• lookup Optional: A character string, naming the lexicon lookup command. Interpreted relative to path/cmd/.
• filter Optional: A character string, naming the output filter to be used. Interpreted relative to path/cmd/.
• no.unknown Optional: Logical, can be used to toggle the "-no-unknown" option of TreeTagger (defaults of FALSE.

You can also set these options globally using set.kRp.env, and then force treetag to use them by setting TT.options="kRp.env" here. Note: If you use the treetagger setting from kRp.env and it's set to TT.cmd="manual", treetag will treat TT.options=NULL like TT.options="kRp.env" automatically.

debug

Logical. Especially in cases where the presets wouldn't work as expected, this switch can be used to examine the values treetag is assuming.

TT.tknz

Logical, if FALSE TreeTagger's tokenzier script will be replaced by koRpus' function tokenize. To accomplish this, its results will be written to a temporal file which is automatically deleted afterwards (if debug=FALSE). Note that this option only has an effect if treetagger="manual".

format

Either "file" or "obj", depending on whether you want to scan files or analyze the text in a given object, like a character vector. If the latter, it will be written to a temporary file (see file).

stopwords

A character vector to be used for stopword detection. Comparison is done in lower case. You can also simply set stopwords=tm::stopwords("en") to use the english stopwords provided by the tm package.

stemmer

A function or method to perform stemming. For instance, you can set SnowballC::wordStem if you have the SnowballC package installed. As of now, you cannot provide further arguments to this function.

Value

An object of class kRp.tagged-class. If debug=TRUE, prints internal variable settings and attempts to return the original output if the TreeTagger system call in a matrix.

Details

Note that the value of lang must match a valid language supported by kRp.POS.tags. It will also get stored in the resulting object and might be used by other functions at a later point. E.g., treetag is being called by freq.analysis, which will by default query this language definition, unless explicitly told otherwise. The rationale behind this is to comfortably make it possible to have tokenized and POS tagged objects of various languages around in your workspace, and not worry about that too much.

References

Schmid, H. (1994). Probabilistic part-of-speec tagging using decision trees. In International Conference on New Methods in Language Processing, Manchester, UK, 44--49.

[1] http://www.cis.uni-muenchen.de/~schmid/tools/TreeTagger/

See Also

freq.analysis, get.kRp.env, kRp.tagged-class

Examples

## Not run: 
# # first way to invoke POS tagging, using a built-in preset:
# tagged.results <- treetag("~/my.data/speech.txt", treetagger="manual", lang="en",
#    TT.options=list(path="~/bin/treetagger", preset="en"))
# # second way, use one of the batch scripts that come with TreeTagger:
# tagged.results <- treetag("~/my.data/speech.txt",
#    treetagger="~/bin/treetagger/cmd/tree-tagger-english", lang="en")
# # third option, set the above batch script in an environment object first:
# set.kRp.env(TT.cmd="~/bin/treetagger/cmd/tree-tagger-english", lang="en")
# tagged.results <- treetag("~/my.data/speech.txt")
# 
# # after tagging, use the resulting object with other functions in this package:
# readability(tagged.results)
# lex.div(tagged.results)
# 
# ## enabling stopword detection and stemming
# # if you also installed the packages tm and SnowballC,
# # you can use some of their features with koRpus:
# set.kRp.env(TT.cmd="manual", lang="en", TT.options=list(path="~/bin/treetagger",
#    preset="en"))
# tagged.results <- treetag("~/my.data/speech.txt",
#    stopwords=tm::stopwords("en"),
#    stemmer=SnowballC::wordStem)
# 
# # removing all stopwords now is simple:
# tagged.noStopWords <- kRp.filter.wclass(tagged.results, "stopword")
# ## End(Not run)