Hunaccent

Accentize Hungarian text. Please note that this libarary does not use a vocabulary and it may generate non-existant words.

Compiling

g++ hunaccent.cpp -std=c++11 -O3 -Wall -pedantic -o hunaccent

Compiling to Javascript

Install Emscripten. Set

#define TARGET_EMSCRIPTEN

at the beginning of hunaccent.cpp by removing the double slash, then run

emcc -std=c++11 --bind -s WASM=0 hunaccent.cpp -o hunaccent.js --embed-file tree

The hunaccent.js and hunaccent.html files are all that is needed, the asset is compiled into the js file by Emscripten.

Usage

$ echo arvizturo tukorfurogep | ./hunaccent
árvíztűrő tükörfúrógép

If you call the binary from a different directory, you need to specify the location of the trees:

$ echo arvizturo tukorfurogep | ./hunaccent/hunaccent hunaccent/tree/
árvíztűrő tükörfúrógép

Deaccentize input

$ echo árvíztűrő tükörfúrógép | ./deaccentize.py
arvizturo tukorfurogep

Accentizing again:

$ echo árvíztűrő tükörfúrógép | ./deaccentize.py | ./hunaccent 
árvíztűrő tükörfúrógép

Note

Hunaccent is still under development, meaning I might change the tree files for better ones anytime.

Citation

Please cite this paper if you use hunaccent in your research:

@article{Acs:2016,
  title={Hunaccent: Small Footprint Diacritic Restoration for Social Media},
  author={Acs, Judit and Halmi, J{\'o}zsef},
  booktitle={Normalisation and Analysis of Social Media Texts (NormSoMe) Workshop},
  year=2016,
}