Lesson 3: Scoring translations and summaries, and their limits
58 min read·14 Oct 2026
Module 9 · Lesson 3 · About 55 minutes · Needs: basic Python; n-grams (Module 4 Lesson 1) and bootstrap intervals (Module 5 Lesson 2) help; this lesson loads its own models
The situation
It is Wednesday 4 November 2026, project week 9. German messages now arrive every day, and a translation model can turn them into English for Ines. Rosa asks one question: "How good is the English?" She asks the same about the thread summaries for her Monday report.
A translation can be right in many wordings and wrong by one word. This lesson scores 14 German translations and the Hartley thread summary four ways (simulated data).