CourseNatural Language Processing: The Complete Course · Module 9: Translation, Multilingual NLP and Evaluating Generation · part 60 of 73
Part 60 · Module 9: Translation, Multilingual NLP and Evaluating Generation

Lesson 1: Machine translation for a four-language inbox

63 min read·14 Oct 2026
Module 9 · Lesson 1 · About 50 minutes · Needs: basic Python; Module 4 (the TF-IDF intent model) and Module 7 Lesson 2 (attention) help; no key needed

The situation

It is Monday 2 November 2026, project week 9. Since Tidewell started delivering to Germany, Spain and France on 1 July, about four in ten messages are in German, Spanish or French. Ines reads Spanish. The German and French ones wait, and the intent model scores worst on them.

One message runs through the lesson: HT48 from the hand-written test set (simulated). "Wo ist meine Bestellung TW-122960? Und ich will das LockSure D-lock umtauschen, der Schlüssel klemmt." Its reference translation: "Where is my order TW-122960? And I want to exchange the LockSure D-lock, the key sticks." Ines labels it order_status, with a warranty fault second. Rosa asks: can a free, local translation model make it routable?

HT48, GERMAN, HAND-WRITTEN Wo ist meine Bestellung TW-122960? ... Schlüssel klemmt. translate MarianMT local, free intent model English Module 4 to Ines order_status
The plan for HT48: a local translation model (amber: a model) turns the German into English so that the English intent model (blue: your system) can route it to Ines (green: correct).

The rest of this part opens with an account

Signing up is free and takes one email. Nothing to pay, and nothing renews.