Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

Building Multilingualism with Claude and Divooka App Platform

Description and use

A static multilingual dictionary built with Claude and Divooka App Platform, offering over a million translations across 108 languages, a public API, and community‑driven correction workflow.

Problem

The existing multilingual lookup site In Different Languages was unreliable and slow, and the author wanted a faster, more reliable dictionary for personal use and to test the Divooka App Platform.

Outcome

A fully functional, backend‑free multilingual dictionary with over a million translation entries, a public API, and an open‑source pipeline that demonstrates AI‑assisted development on the Divooka App Platform.

Workflow

  • 1. Extract English entries with translation tables from Wiktionary JSONL via kaikki.org.
  • 2. Select common words using wordfreq and prioritize coverage.
  • 3. Identify missing word‑language pairs and generate translation jobs for Claude with meaning and part of speech.
  • 4. Validate machine translations against hidden Wiktionary control samples, flag uncertain entries, and prefer Wiktionary when available.
  • 5. Merge results, organize by word and language, and publish compact JSON files with metadata.
  • 6. Host data on GitHub and serve via jsDelivr CDN; build a plain‑HTML/JS frontend that fetches needed JSON.
  • 7. Expose a public JSON API for developers and provide correction links to GitHub issues.

Evidence

Unfortunately, the website frequently becomes unavailable or unreasonably slow.

Source

The entire application was built with Claude through Divooka App Platform's MCP connector.

Source

We started with the English Wiktionary, using structured extracts provided by kaikii.org through Wiktextract.

Source

The resulting dataset contains 9,852 English words and phrases across 108 languages, representing over a million word-language translation entries.

Source

All translation data is stored as JSON in the GitHub repository and served through the jsDelivr CDN.

Source

We also preserve the translation sources and publish machine-generated entries as reviewable TSV files.

Source

  • Record date: 2026-10-22T20:11:22.327Z