Skip to content

first-session.md

Latest commit

 

History

History
167 lines (120 loc) · 4.75 KB

first-session.md

File metadata and controls

167 lines (120 loc) · 4.75 KB

Tutorial: Your First Translation Session

This tutorial walks you through a complete translation session — from setup to submitting your results and seeing them on the dashboard.

Prerequisites

Make sure you've completed Getting Started:

  • gh auth login done
  • Repo cloned and uv sync run
  • uv run qebench doctor passes all checks

Step 1: Update

Always start by pulling the latest code and data:

uv run qebench update

This runs git pull --rebase and uv sync so you have everyone's latest contributions and any CLI updates.

Step 2: Check Your Dataset

uv run qebench stats

You should see the coverage panel and domain table. The dataset is pre-seeded with 314 economics and math terms — enough to start practicing immediately.

Step 3: Start a Short Session

Let's try a 3-entry session with basic-level terms:

uv run qebench translate -n 3 --difficulty basic

Your GitHub username is detected automatically from gh auth.

Step 4: Translate Each Entry

For each entry, you'll see a panel like:

╭──────────────── TERM ────────────────╮
│                                      │
│  inflation                           │
│                                      │
│  Context: "The rate of inflation     │
│  determines the real interest rate." │
│                                      │
│  term-001 · economics · basic        │
╰──────────────────────────────────────╯
? Your translation (中文):

Terms with context sentences (populated by qebench update) show an example from a QuantEcon lecture to help you choose the right translation.

Type your Chinese translation and press Enter.

Step 5: Rate Your Confidence

After typing your translation, you'll be asked how confident you are:

? How confident are you?
  1 — Guessing
  2 — Uncertain
❯ 3 — Reasonable
  4 — Confident
  5 — Very confident

Pick the level that matches your certainty. You can also add optional notes (context, reasoning, alternative phrasing) — just press Enter to skip.

Step 6: See the Reference

After you commit your answer, the reference translation is revealed with a similarity score:

╭────────── Reference ──────────╮
│  Your answer:  通货膨胀        │
│  Reference:    通货膨胀        │
│  Similarity:   100%              │
│  Matches the reference.          │
╰───────────────────────────────╯

If your translation differs (below 85% similarity), you'll be asked why:

? Why does your translation differ?
  Formal/written register (书面语)
  Informal/spoken register (口语)
❯ Regional preference
  Contextual — depends on usage
  Abbreviation or shorthand
  Alternative technical term
  Other (explain in notes)

? Notes (optional — explain further):

This is the most valuable data you can provide. Chinese has formal and informal representations, regional variants, and context-dependent translations. Documenting why translations vary helps us improve the translator.

Step 7: Session Summary

After completing all entries, you'll see:

╭─── Session Summary ───╮
│  Completed:  3/3       │
│  XP earned:  +30       │
│  Total XP:   30        │
╰────────────────────────╯

Each completed translation earns 10 XP. Your results are saved locally to results/translations/{your-username}.jsonl and your XP to results/xp/{your-username}.json.

Step 8: Submit Your Results

Push your work to GitHub so it appears on the dashboard:

uv run qebench submit

This pulls the latest changes, commits your results, and pushes — all in one command.

╭────── Submitted as alice ──────╮
│  ✓ 1 result file(s)           │
╰────────────────────────────────╯

Once pushed, the CI workflow rebuilds the dashboard with your activity and XP.

Step 9: Try Different Filters

Focus on a specific domain:

uv run qebench translate -d mathematics

Or crank up the difficulty:

uv run qebench translate --difficulty advanced -n 10

Next Steps