Module 1: Text as Data and Rules
1 min read·14 Oct 2026
What you need first
- Basic Python: strings, lists, dictionaries, loops and functions (the separate Python course covers them).
- A Google account for Colab and Drive. No NLP, and no model key: this module's only hosted call, in Lesson 1, is optional.
What you will build
The start of the tidewell-inbox repository: the shared data as inbox/data.py, a tested text cleaner, field extraction, redaction and a keyword rule router, the baseline (0.878 on template messages, 0.438 on HARD_TEST) that every later model must beat.
Lessons
| File | Title | Minutes | Depth |
|---|---|---|---|
01-what-nlp-is.md | Lesson 1: What NLP is, and the inbox we will fix | 45 | Essential |
02-characters-encodings-and-normalisation.md | Lesson 2: Characters, encodings and clean inbox text | 60 | Essential |
03-regular-expressions-for-messages.md | Lesson 3: Regular expressions that find order numbers and dates | 45 | Essential |
04-rules-in-production.md | Lesson 4: Rules in production: safe, tested and measured | 60 | Essential, Recommended |
Extras
05-module-lab.md, Lab: The inbox package and its rule baseline (75 minutes): the repository in your Drive,%%ensure, the data, cleaner, rules and tests, and the baseline score.06-incident-file.md, Incident: A name key that changed twice (30 minutes): Spotify's 2013 username hijacking, replayed on Tidewell's customer names, and the idempotence test that catches it.