In vanatteveldt/frogr: R client for the frog tagger and parser for Dutch

Calling frog from R

Frog is a lemmatizer and dependency parser for Dutch which can also be run as a server. This package contains functions for connecting to a frog server from R and creating a document-term matrix from the resulting tokens. Since this yields a standard tm term-document matrix, it can be used e.g. for corpus analysis, topic modeling, or machine learning using RTextTools

See http://ilk.uvt.nl/frog/ for more information on Frog.

Installing and running the frog server

The frog daemon (server) needs to be running before you can this package. See http://ilk.uvt.nl/frog/ for documentation and installation instructions.

To install frog on debian/ubuntu you can use apt:

$ sudo apt-get install frog frogdata ucto

To run the frog server on port 9772, use:

$ frog -S 9772

If you only want to pos-tag and lemmatize, you can skip the parsing and morphological analysis to speed up the analysis and conserve memory:

$ frog --skip=acpm

Installing frogr

frogr can be installed directly from this github repository using devtools:

if (!require(devtools)) {install.package("devtools"); library(devtools)}
install_github("frogr", username="vanatteveldt")
library(frogr)

If devtools is unavailable (e.g. on Windows), you can also copy the file frog.r and source it directly. In that case, make sure the packages tm, Matrix and zoo are installed.

Calling frog

The function call_frog calls the frog server with a give text and results a data frame:

text = c("Mijn kat Toby heeft nooit van andere katten gehouden.",
         "Maar andere katjes houden wel van hem!")
tokens = call_frog(text, host="localhost", port=9772)
head(tokens)

Note that if you run frog with the --skip= argument, some columns will only contain NA values. The sentence and majorpos columns are not produced by frog but included here for convenience. majorpos is simply the part of the POS tag before the first parenthesis.

Creating a document-term matrix

To create a document term matrix from the frog output (or in fact from any list of tokens), you can use the create_dtm function:

m = create_dtm(tokens$docid, tokens$lemma)
as.matrix(m)

Of course, you can also first select to e.g. only keep nouns and verbs:

subset = tokens[tokens$majorpos %in% c("N", "WW"), ]
m = create_dtm(subset$sent, subset$lemma)
as.matrix(m)

As you can see, all forms of cat (kat, katten, katjes), love (houdt, houden), and have (heeft) are properly lemmatized.

vanatteveldt/frogr documentation built on May 3, 2019, 2:59 p.m.