Skip to content

Latest commit

 

History

History
64 lines (42 loc) · 1.98 KB

README.md

File metadata and controls

64 lines (42 loc) · 1.98 KB

Flect

Robust Multilingual Statistical Morphology Generation Models

Authors: Ondřej Dušek and Filip Jurčíček

Institute of Formal and Applied Linguistics, Charles University in Prague

About

Flect is a tool based on Python and Scikit-Learn that learns morphological inflection patterns from corpora.

Use any morphologically annotated corpus to have the system learn how to automatically obtain inflected word forms from lemmas and morphological features.

The system is able to inflect even previously unseen words by using lemma suffixes as features and predicting “edit scripts” that describe the difference between the lemma and the form.

Flect schema

Basic usage

  • Prepare data using the bin/conll2arff.py script.

  • Train your models and save them using bin/train.py.

    • setup the models' configuration via Python code as shown in configs/.
  • Test the performance of your models using bin/test.py.

  • To easily access Flect as a library from your program, use the SentenceInflector class located in the flect.flect module.

Additional functions

  • You may generate further features (concatenations etc.) using the bin/combine_features.py script.

  • Data statistics may be obtained from bin/get_data_stats.py, bin/get_feat_list.py and bin/select_errors.py.

License

Distributed under the Apache 2.0 license. See LICENSE for more information.

Please cite the following paper if you use this software in your scientific works:

  • Ondřej Dušek, Filip Jurčíček: "Robust Multilingual Statistical Morphological Generation Models", in: ACL Student Research Workshop, Sofia, 2013.

The paper which describes the inner workings of the software and our experiments done with it, is available here for download.

You may contact the authors at odusek * ufal.mff.cuni.cz or through GitHub in case of bugs, comments, or questions.