Skip to content
This repository has been archived by the owner on May 10, 2023. It is now read-only.

Latest commit

 

History

History
187 lines (135 loc) · 8.18 KB

README.md

File metadata and controls

187 lines (135 loc) · 8.18 KB
Raw

Common Voice Sentence Collector

The Sentence Collector is part of the Common Voice project. Its purpose is to provide a tool for contributors to upload public domain sentences, which then can get reviewed and are exported to the Common Voice database. Once imported they will show up for contributors on Common Voice to read out aloud.

Get involved

  • Fork the project and test that you can run the environment locally following the instructions below.
  • Is everything working as expected? If not, submit a new issue.
  • Review the pending issues in the project.
  • Create a new PR to fix any of the existing issues in the project.
  • To add or adjust normalization, validation and cleanup for a language see VALIDATION.md and CLEANUP.md. Cleanup is optional and should only be used in rare cases.

Prerequisites

Architecture

Diagram

To edit this diagram, load the architecture.svg in the docs of the repository into diagrams.net and then save the updated version back into the repository like any other file changes you'd make.

Local Development

Start the MySQL database:

docker-compose up

Now we can install the dependencies:

npm run install:all

Start the server in a new terminal window. For this you'll need parameters from Auth0 to make sure that the login works, see below for a short tutorial.

cd server
npm run db:migrate
env AUTH0_DOMAIN=yourusername.eu.auth0.com AUTH0_CLIENT_ID=fromauth0 AUTH0_CLIENT_SECRET=fromauth0 SESSION_SECRET=somerandomvalue npm start

Finally, you can start the frontend in another terminal window. Please make sure that you're in the root directory of the repository.

cd web
npm start

The sentence collector is now accessible through http://localhost:3333. We're serving the frontend code through the NodeJS app to make Auth0 work locally. However the frontend code changes are still rebuilt, so you can reload the tab to see changes.

Seed

If you want to fill up your local database with a lot of sentences and votes, you can run the following script. This command will run for a long time, as it creates millions of entries in the database. For normal development this might not be necessary, however this can be very useful for performance measurements.

npm run db:seed

However note that this only works on a freshly created database for now, to keep the complexity with foreign keys low. If somebody wants to adjust the seed script to be more dynamic and not rely on hardcoded keys, we are happy to accept a PR for this.

Working with localizations / languages

We are using the same languages as Common Voice is. Therefore we rely on the information Common Voice stores in its repository. The locales JSON files in this repo only contain example data. These files get updated when deploying Sentence Collector.

If you want to have all the translations available locally, you can run npm run pull-cv-translations in the root folder. Note that even after pulling this information, not all strings will be translated. We are not fetching the complete English translation file, therefore the language selection on development will be broken. This is not perfect, but for now expected behavior. This is a drawback of the current infrastructure, as we do not want the complete English translation file from Common Voice, as we keep the Sentence Collector EN FTL strings under version control. We do not want to risk to push the CV strings twice, therefore we do not touch the English file. If you want to test it, use a different UI language after running npm run pull-cv-translations.

Getting required parameters from Auth0

  1. Create an Auth0 account. There is a free tier and that will be enough for local development.
  2. Click "Applications" from the dashboard. Create a new one, or use the default application.
  3. Go to "Applications" and click on the Settings icon next to your application.
  4. Add http://localhost:3333/sentence-collector/callback to the "Allowed Callback URLs" list.
  5. You can add more login options to your app from the "Connections" tab, but you don't have to
  6. These required domain, client ID and client secret are found in the same Settings tab as the "Allowed Callback URLs". Copy those into your run command from above.
  7. You will now be able to create a new user by clicking on "Login" and then switching over to the "Sign Up" tab on the login dialog. You don't need to use a valid email address.

Running automated tests

Backend

Running the unit and integration tests:

cd server
npm t

Frontend

Running the unit tests:

cd web
npm t

If you need to update the snapshots, run npm t -- -u, but make sure that the new snapshots indeed match the expected outcome.

Path special cases

  • In production the app is mounted at /sentence-collector
  • Therefore we need to make sure that we're not requesting resources on /
  • This however leads to a weirder case locally, as we're faking this relationship.
  • We do not want to build the base path into the image, which does not leave us many possibilities. Injecting it into the running container is a possibility, but makes it quite complicated for the frontend calls.
  • Therefore we are serving the frontend on / locally, and the API lives at /sentence-collector
  • With that we have the same relative calls as we have in production
  • Apart from having a prefix that is either empty or not, we do not have any other special casing

Sentence Export to Common Voice

You don't need to manually export the sentences from the Sentence Collector to the Common Voice repository. This is done automatically once a week: https://github.com/common-voice/sentence-collector/actions?query=workflow%3Aexport

Localization Process

The localization for the Sentence Collector happens in the Common Voice project. You can learn more about that in the LOCALIZATION documentation.

Useful queries

Get all approved sentences

SELECT
      Sentences.id,
      Sentences.sentence,
      Sentences.localeId,
      SUM(Votes.approval) as number_of_approving_votes
    FROM Sentences
    LEFT JOIN Votes ON (Votes.sentenceId=Sentences.id)
    GROUP BY Sentences.id
    HAVING
      number_of_approving_votes >= 2;

Get not decided sentences

SELECT
      Sentences.id,
      Sentences.sentence,
      Sentences.localeId,
      SUM(Votes.approval) as number_of_approving_votes,
      COUNT(Votes.approval) as number_of_votes
    FROM Sentences
    LEFT JOIN Votes ON (Votes.sentenceId=Sentences.id)
    GROUP BY Sentences.id
    HAVING
      number_of_votes < 2 OR # not enough votes yet
      number_of_votes = 2 AND number_of_approving_votes = 1; # a tie at one each

Get all decided

SELECT Sentences.*
    FROM Sentences
    LEFT JOIN Votes ON (Votes.sentenceId=Sentences.id)
    GROUP BY Sentences.id
    HAVING
      COUNT(Votes.approval) >= 2;

Get all rejected

SELECT Sentences.*
    FROM Sentences
    LEFT JOIN Votes ON (Votes.sentenceId=Sentences.id)
    GROUP BY Sentences.id
    HAVING
      (
        COUNT(Votes.approval) = 3 AND
        SUM(Votes.approval) < 2
      ) OR (
        COUNT(Votes.approval) = 2 AND
        SUM(Votes.approval) = 0
      );

API

You can find the API documentation for the public endpoints in the OpenAPI spec.