Translation Project Brainstorming Meeting Agenda and Notes

5 minutes

Intro to Translation Tickets

• https://archivesspace.atlassian.net/jira/software/projects/TD/boards/57

  • https://archivesspace.atlassian.net/browse/TD-21

    • Definitely want to provide translations for the TechDocs.

    • Some stuff in the TechDocs website can be translated as part of the web interface - but not the content.

      • App interface is auto-translated.

    • Add top level directories for each language.

      • root == default which is english (could have a en folder for english)

    • Content is different. ASpace is a key-value pair with the locales files.

      • english keys, values would be translated.

      • Content will need to be

  • https://archivesspace.atlassian.net/browse/TD-54

• Languages identified:

  • https://archivesspace.atlassian.net/browse/TD-55

  • https://archivesspace.atlassian.net/browse/TD-56

  • https://archivesspace.atlassian.net/browse/TD-57

  • https://archivesspace.atlassian.net/browse/TD-58

  • Need to fresh the language list to the list of supported translations. Ask Jessica what is the list and that will be our target.

30 minutes

Brainstorming ideas on project

• If LLMs - be hands on with it.

  • Going section by section, giving context to the LLM each time.

  • Then ask community fluent speakers to review the work of the LLM

    • Send the PR links to community members to review.

  • If we don’t get community members reviewed, we need to note translations done by LLMs

• What about weblate?

  • Sits in between individuals and repo.

  • Brian suggests weblate is overkill for this project.

    • PR is the source of the translation.

• For ASpace core, what has been translated?

  • Nothing is 100% translated in a language in ASpace for other languages. But varying degrees of translation - all dependent on individual’s work.

  • EN, FR, SP, DE, UK, JP

• Work is going to be file-based.

  • Start with files being divided.

  • TD-21: Translate the Tech Docs by brianzelip · Pull Request #320 · archivesspace/tech-docs

• Requirements:

  • Make sure to match the languages that are officially supported

  • Document the process for how to translate the content of the docs

    • Step-by-step for branching, dividing work for pages, using LLM for translating, generating markdown file, posting PRs for review.

      • Could also use a secondary LLM to review the first’s LLM’s translations.

  • Secondary goal is adding the other languages in ASpace/locales folder that are not officially supported, but are working on.

• Do we have a preferred LLM? Consistent Model? Model ideal for translation?

  • Chances are LLMs are aware of ArchivesSpace.

  • Could also review LLM’s first pass with a different LLM for review.

• Exclude translations for specific parts of pages:

  • Code blocks

  • Folder names in the ASpace core code directory - ex. ‘backups’ directory is blah blah == ‘backups’ needs to stay English.

  • ASCII art ( - ex. docker.md example of file structure)

  • TO INVESTIGATE:

    Internal links - these same page links to subheaders in same page or different pages in techdocs will be translated - so need to review these too.

  • NOTE: use Branch Preview URL for PR’s to review - this URL will remain the same.

• Brian uses Cursor in VSCode - suite of LLMs.

• LLM’s?

  • Co-pilot for Corey

  • Co-pilot for Laura

  • DeepL - also good for translators

  • Alex - not sure what they have access to.

20 minutes

Next steps for project

• Save project for next term to start doing actual work, but can start prepping documentation and step-by-step process with Brian’s help.

10 minutes

Final review of tasks

• TODOs: TechDocs team members record what LLMs do they have access to/permission to use?

  • Ok to mix and match tools from Brian’s opinion - most important to use what you have.

• TODO @Brian Zelip : refresh the language list to the list of supported translations. Ask Jessica what is the list and that will be our target.

• TODO @Brian Zelip : start developing step-by-step process for this translation work.

• TODO @Corey Schmidt : start preparing project documentation with help from Brian.

  • Determine what other questions we have:

    • Community review? LLM secondary review?

    • Designation of work for translation? Per page? Per language? Other?

    • What LLM’s to TechDocs team members have access to?

    • Build in exclusions for things like: code blocks, folder structures?

    • How to handle internal links?

    • How will LLM’s tolerate spelling/grammar errors? Will we need to review all TechDocs contents before submitting to LLM for translation?