Norma

Norma is a tool for automatic spelling normalization of non-standard language data. It uses a combination of different normalization techniques that typically require training data (= a list of manually normalized wordforms) and a target dictionary (= a list of valid wordforms in the target language).

Besides this README, see also the User Guide for more information on how to use Norma.

If you have any questions, suggestions, or comments, please contact one of the authors:

• Marcel Bollmann ([email protected])
• Florian Petran ([email protected])

License

Norma is licensed under the GNU Lesser General Public License (LGPL) v3.

Usage via Docker

If you don't want to compile Norma and its dependencies from scratch, you can try to use it via a Docker image. To do so, install Docker (e.g. via your system's package manager) and run:

docker run -v $(pwd):/home mbollmann/norma

This should download the Docker image and display Norma's help output. You can now add command-line arguments the same way as with the normalize binary; for example, if you cloned this repository locally, you should be able to run:

docker run -v $(pwd):/home mbollmann/norma -s -c doc/example/example.cfg -f doc/example/fnhd_sample.txt

To use the norma_lexicon command with the docker image, run:

docker run -v $(pwd):/home --entrypoint norma_lexicon mbollmann/norma <args...>

Make sure that the files you specify exist within your current working directory.

Dependencies

• Needed for compilation:
  • GCC >= 4.9
  • CMake >= 2.8.10
  • Boost >= 1.54
    • in particular these libraries: Filesystem, Program Options, Regex, System, Test
  • pkg-config
  • gfsm >= 0.0.16-1 and gfsmxl >= 0.0.15, available from http://kaskade.dwds.de/~moocow/mirror/projects/gfsm/
  • GLib >= 2.0
• Optionally:
  • ICU >= 1.49
  • Doxygen (for generating the documentation)
  • Python 2 >= 2.7 and Boost::Python (for Python bindings/embeddings)

How to do an out of source build

mkdir build
cd build
cmake <pathtosource>
make

There is a test suite included that can be run with make test (or make check, which reveals more detailed error messages), as well as API documentation that is generated via make doc (requires Doxygen).

Configuration options (for CMake)

• String implementation (default: ICU if available): -DSTRING_IMPL=(ICU|STD)
  • ICU - use ICU unicode strings
  • STD - use STL string - requires no additional library
• Build type (default: Release): -DCMAKE_BUILD_TYPE=(Debug|Release):
• Install prefix (default: /usr/local/) -DCMAKE_INSTALL_PREFIX=<prefix>
• To make Python bindings/embeddings (default: disabled), set -DWITH_PYTHON=TRUE

Other platforms

• Norma was developed on Linux and will not work on non-Unix platforms.
• Compilation with clang will probably work but is untested.