Translation Linter

GitHub action for auto-translating l10n files back into English for easier review. Also automatically checks for potentially offensive language in translation files, as well as translation similarity.

Auto-translations are provided by Argos Translate, profanity checking provided by (Alt) Profanity Check, and similarity checking is done by spaCy.

At the moment this action supports JSON translation files like .arb, as well as gettext formats like .po and .pot.

Usage

Example Workflow File

Add the following as a GitHub workflow file (e.g. .github/workflows/example-action.yaml):

name: YOUR_WORKFLOW_NAME
on:
  workflow_dispatch:
  pull_request:
  push:
    branches: [main]

jobs:
  YOUR_JOB:
    name: YOUR_JOB_NAME
    runs-on: ubuntu-latest
    steps:
      - name: Checkout
        uses: actions/checkout@v4
      - name: YOUR_STEP_NAME
        uses: ashuntu/translation-linter@main
        with:
          # Change this to the directory containing your translations
          # REQUIRED
          translation-dir: "translations"
          # RegEx file mask to match translation files
          # The first group match () should contain the language code
          # OPTIONAL
          file-mask: "app_(.*)\\.arb"
          # File path to original, untranslated file for similarity checking
          # OPTIONAL
          template: "translations/app_en.arb"

You may also customize the thresholds for profanity/similarity warnings. See all input options in action.yml.

Testing Locally

You can test this action locally by doing the following:

1. Clone the repository
2. Run pip install -r requirements.txt
3. Run python main.py with options

   # For example...
   python main.py --source example-translations --mask "example_(.*)\\.arb" --template example-translations/example_en.arb

Output

See the most recent action run from this repo for an example of the output this action generates.

Profanity warnings are on a scale of 0 to 1, where 0 is definitely no profanity and 1 is most likely profanity. Similarity warnings are on a scale of 0 to 1, where 0 is completely similar, and 1 is not at all similar.

Example output screenshots