RNNTagger

RNNTagger - a Neural Part-of-Speech Tagger

The RNNTagger is a tool for annotating text with part-of-speech and lemma information. It comes with pretrained parameter files for many languages. RNNTagger was implemented in Python using the Deep Learning library PyTorch.

Compared to TreeTagger, the pros of RNNTagger are

higher tagging accuracy
lemmatizes all tokens

The cons are:

slower
requires PyTorch
requires a GPU for improved speed
larger parameter files
Lemmas of unknown tokens are guessed and are therefore not always correct

Download

This software is freely available for research, education and evaluation. For commercial and other licenses, please contact the developer via the email address at the bottom of the page.

Please read the license terms, before you download the software! By downloading the software, you agree to the terms stated there.

The following steps are required to install RNNTagger on Linux:

Install PyTorch and CUDA (the latter is only needed for GPU support)
Download the RNNTagger package (2.6 GB)
Unzip the RNNTagger.zip archive

Now, you can open a command-line shell, change to the newly created directory RNNTagger, and enter the commands:

> echo "This is a test." > test.txt
> cmd/rnn-tagger-english.sh test.txt
This will produce the output:

This	DT	this
is	VBZ	be
a	DT	a
test	NN	test
.	.	.

Currently supported modern languages: Bulgarian, Catalan, Czech, Danish, Dutch, English, Estonian, Finnish, French, Galician, German, Greek, Hindi, Hungarian, Icelandic, Italian, Korean, Norwegian, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swahili

Currently supported ancient languages: Latin, Middle Dutch, Middle English, Middle High German, Old French, Old Greek, Old Icelandic, Old Italian

The tagger package contains a README file with further information on the tagger and the parameter files.

Please send questions, comments, suggestions and bug reports to Helmut Schmid at LastName@cis.lmu.de.