Lezvaget (Լեզվագետ)

A web app for Eastern Armenian language tooling — starting with spellchecking, with a verb conjugator planned next.

Status

Early scaffold. Currently working:

• Rails app with PostgreSQL
• SpellChecker service wrapping Hunspell via ffi-hunspell, using the vendored hy_AM (Eastern Armenian) dictionary

Not yet built: web UI, conjugation engine, custom-word extension table.

Built with

• Ruby on Rails
• PostgreSQL
• ffi-hunspell for spellchecking
• RSpec for testing

Dictionary

The Hunspell dictionary (lib/dictionaries/hy_AM/) is vendored from martakert/hyspell (CC0-1.0). See lib/dictionaries/hy_AM/SOURCE.md for pull date and details.

Dependencies

Docker

Setup

docker compose run --rm -- web
# inside container:
bundle install
bin/rails db:create db:migrate
bin/rails console

Spellchecker

SpellChecker.new.check("տուն")  # => true
SpellChecker.new.check_text("Բարև ինչպես ես") 
# => [{word: "Բարև", valid: true}, {word: "ինչպես", valid: true}, {word: "ես", valid: true}]

Conjugator

# seed some lemmas
Lemma.create!(word: "գրել", part_of_speech: "verb", conjugation_class: "e_conjugation")
Lemma.create!(word: "խաղալ", part_of_speech: "verb", conjugation_class: "a_conjugation")
# conjugate verbs
Conjugator.new.present_tense(Lemma.find_by(word: "գրել"))
Conjugator.new.present_tense(Lemma.find_by(word: "խաղալ"))

TODO

• Web UI for pasting text and highlighting misspelled words
• CustomWord model for user-added proper nouns/names
• Conjugator service for generating verb paradigms
• Rule-based guessing for regular verbs (def guess_class(word)), database only for exceptions, should work without saying what the conjugation_class is
• Public deployment