The Oslo-Bergen Multitagger for Norwegian Bokmål and Nynorsk

The Python version was written by Michał Kosek (Text Laboratory, University of Oslo). It has been tested with Python 2 and Python 3.6.6. Using version >= 3.7 may cause problems with word forms containing regular expression special characters.

The code is partially based on the Perl Multitagger written by Lars Jørgen Tvedt (Dokumentasjonsprosjektet, University of Oslo).

The output of the multitagger reproduces the format used by the Lisp Multitagger written by Paul Meurer, University of Bergen. To produce an output that most resembles the Lisp version, use -compat in the command line.

The compound analyser has been written by Michał Kosek, using rules formulated by Janne Bondi Johannessen and Helge Hauglin (1998).

The code is freely available under the MIT licence. The associated grammatical information comes from Norsk Ordbank, National Library of Norway, and is available under the CC-BY licence.

Bug reports and comments can be directed to tekstlab-post@iln.uio.no.

Johannessen, J. B., & Hauglin, H. (1998). An automatic analysis of Norwegian compounds.

Troubleshooting

The multitagger will refuse to process invalid UTF-8:

$ ./mtag.py textfile
Traceback (most recent call last):
[...]
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xf8 in position 1252: invalid start byte

It usually means that the input file has different encoding, and needs to be converted to UTF-8 before processing. However, if you are sure you want to process the file with invalid UTF-8 as it is, call the tagger as follows:

$ PYTHONIOENCODING=utf-8:surrogateescape ./mtag.py < textfile

Note that this requires Python 3, and that the input needs to be provided by redirecting stdin with <.