This repository contains the scripts required to reproduce the results published in An unsupervised method for weighting finite-state morphological analyzers as an oral paper for LREC 2020.

• lttoolbox (version: 3.5.1)
• hfst (version: 3.15.2)

• Clone this repository
• Clone apertium-weighting-tools: git clone https://github.com/apertium/apertium-weighting-tools.git (go to branch: evaluation)
• Clone the following repositories at the same directory as that of the evaluation repository:
  • apertium-eng: git clone https://github.com/apertium/apertium-eng.git (commit SHA1: 076aaa3920eb78432d431541672219d455198f86)
  • apertium-kaz: git clone https://github.com/apertium/apertium-kaz.git (commit SHA1: 51fc4e532eb61e589a796116ab973069f26b7145)
  • apertium-hbs: git clone https://github.com/apertium/apertium-hbs.git (commit SHA1: a34d0d5eff6b03239c5bcf9fd49207645444f485)
• Make sure all the repositories share the same root directory

|
|-- apertium-weighting-tools-evaluation/
|-- apertium-weighting-tools/
|-- apertium-eng/
|-- apertium-kaz/
|-- apertium-hbs/

1. Add the apertium-weighting-tools and apertium-weighting-tools/eval to the PATH variable. e.g: export PATH="$PATH:ABSOLUTE_DIR_TO_WEIGHTING_TOOLS:ABSOLUTE_DIR_TO_WEIGHTING_TOOLS/eval"
2. Choose a language
3. cd to the directory of this language
4. Run the script
5. The results will be generated in the LANG/results directory